Overview and Background

Soil-water interactions are fundamental in a number of environmental processes and play a pivotal role in flood management, plant growth, and nutrient cycling. However, these interactions are highly complex and we currently rely on computationally intensive process-based models to help understand these processes and predict their influence on ecosystem services. With recent advances in satellite imagery and sensing, soil moisture data and other relevant data products are now available at spatial and temporal scales suitable for enhancing these models. However, integrating large volumes of data with these complex models is challenging. This studentship focuses on taking advantage of new exascale computing approaches to facilitate data assimilation, exploring how the fusion of big-data with soil-water process models can help unlock new insights and understanding.

Teaser Project 1: Improve process-based model representation of the long-term effects of extreme weather on soil carbon and nutrient cycling through remote sensing data assimilation

The lack or over-abundance of water can have large effects on plants, especially on annual crops where water conditions can severely affect the plant’s growth and survival. With changing water patterns and increasing frequency of heat waves and extreme rainfall events, the effects on plant productivity are expected to be large, and there will be knock-on effects for soil carbon storage and nutrient cycling in the longer-term. Remote sensing offers many data products that can provide us with data-based insights into plant productivity and soil moisture conditions. However, remote sensing of soil carbon is much more difficult, and understanding of the long-term response to changes in plant productivity still requires process-based models. In this project we will experiment with combining remote sensing data and process-based models to better understand the long-term effects of extreme weather on soil carbon and nutrient cycling.

Methods and PhD Pathway:

• The process-based model N14CP, which simulates plant-soil carbon, nitrogen and phosphorus cycling and export of dissolved nutrients to waterways will be adapted to assimilate (Gross or Net) Primary Productivity (GPP and NPP) and soil moisture data remote sensing products.
• The student will explore a range of approaches: from direct insertion to machine learning methods. The first teaser will begin with NPP as this has a direct proxy in the model. Freely available datasets for example from MODIS and SMAP that match the spatial resolution of the model will provide a starting point.
• To develop this path into a full PhD: multiple methods for assimilation will be explored; two-way learning between data and models will be considered; and methods for making data-assimilation real-time will be explored, with the use of exascale/GPU computing, helping move towards a digital twin.

Teaser Project 2: Estimate the contribution of soils to mitigating or increasing flood risk in a case study catchment by combining remote sensed soil moisture data and hydrological models

Antecedent soil saturation conditions can play a significant role in mitigating or increasing flood risk. If the soil has significant water held in storage, then its ability to act as a storage during times of high rainfall reduces.

Soil moisture is an important component in semi-distributed or distributed hydrological models, however, it is not routinely updated dynamically throughout the process of a simulation. With newly available satellite observations soil moisture is now available at good resolution temporally and spatially, such that is could be used to improve flood routing and water balance within catchment hydrological models.

Combining remote sensing and soil water probes offers an opportunity to develop real-time estimation for soil water, across catchments. The student will explore different approaches to data assimilation and upscaling to the catchment scale.

Methods and PhD pathway:

• Combining soil moisture estimates into hydrological and eventually hydraulic modelling for flood inundation estimation will entail significant challenges; firstly in the assimilation of data, secondly through the computational burden, and thirdly through the coupling of models in an online dynamic manner.
• Each of these challenges requires different innovative study and a range of methods.
• The student will be able to pick one or more of these three challenges to explore and develop into a full PhD, should they choose this pathway.
• For example, the coupling between hydrology requires the exploration of different coupling approaches, and will require consideration of processes such as the Basic Model Interface (BMI).

References and Further Reading

1. Plant-soil UK scale model example: Yumashev et al. 2022 Environ. Res. Lett. 17 114054 (click here)
2. Hydrological data assimilation example: Li, H., Huang, Y., Qi, Y. et al. 2024. Water Resources Management 38, 4933–4953 (click here)
3. Information on N14CP Model (click here)

Overview and Background

Understanding biodiversity is crucial to predict the impacts of global change and inform conservation strategies. One aspect of biodiversity that has been largely overlooked is biological colour, due to its complex quantification compared to other facets of biodiversity such as species richness. We therefore have little understanding of why species are the colours they are, its role in mediating interactions with other species (reproduction, competition, predation) and with the environment (e.g., thermoregulation), and the implications of environmental change on this role (e.g., effect of changing background on camouflage ability). One reason research in this area is limited is that quantification of colour and pattern is conceptually challenging and computationally intensive. To advance our understanding of biological colour, new more efficient techniques must be developed for both its quantification, visualisation and interpretation.

Methodology and Objectives

Biodiversity is entering the realms of ‘big data’ with the rapid growth in Citizen Science, emerging sensors for more automated monitoring, and progress towards digital twins. These approaches are generating large image datasets in near-real-time. Leveraging recent advancements in machine learning (ML) and artificial intelligence (AI), we can unlock unprecedented insights into these datasets. One area of interest is in enabling detailed analysis of biological colouration and patterning. State-of-the-art techniques such as convolutional neural networks (CNNs) and vision transformers (ViTs) can allow precise quantification of colour distributions across species and ecological communities. Through such approaches, it becomes possible to monitor and predict the effects of environmental change on biological colouration, informing conservation strategies and advancing our understanding of ecological processes.

Teaser Project 1: Have community colourscapes shifted over time in response to altered environmental conditions?

Objectives:

1. Quantify colour across insect species from citizen science images.
2. Reveal the diversity of colour and pattern across UK insect species.
3. Follow-on. Create new methods to determine assemblage-scale colourscapes.
4. Follow-on. Determine if and how insect assemblage colouration has changed over time and space, and explore the implications of that change for ecosystem function

This project draws on the concept of bioacoustic “soundscapes” to explore assemblage-scale shifts in the context of environmental change. For the teaser project, the student will apply CNNs and ViTs to quantify the colour of multiple insect species through access to two main image datasets: (1) images submitted via the iRecord citizen science platform, which includes >5 million photos spanning >20,000 UK species; and (2) standardised images of nocturnal insects captured by the UKCEH Automated Monitoring of Insects (AMI) sensor network in Central America, Africa, and Asia involve 40 devices generating ~2Tb of image data per year. This quantification will provide the foundation for a larger project coupling these colour quantifications with temporal assemblage data derived from decades of species observations across the UK (UKCEH Biological Records Centre; GBIF). This integration would enable analysis of how colourscapes – the frequency and distribution of colour across co-occurring species – have changed nationally over time, and the ecological implications of such changes. To engage conservation stakeholders, the student could later develop interactive geospatial visualisations of colour changes over time. Once calibrated, the project could be expanded to other regions and taxa, particularly where AMI sensors are deployed, and/or with coral reef fishes and birds leveraging supervisory expertise. 

Teaser Project 2: Does habitat degradation disrupt the evolutionary match between animal colouration and its background?

Objectives:

1. Quantify coral reef background colouration using computational methods applied to 3D photogrammetry data.
2. Reveal the relationship between habitat colour diversity and reef degradation.
3. Follow-on. Assess potential mismatch between reef fish colouration and habitat colour in degraded versus healthy reefs, and its implication for ecological function.

This project explores how coral reef habitat colour diversity is affected by environmental degradation using 3D models built from photogrammetry. The teaser project will build towards a broader exploration of how environmental change impacts the ecological and evolutionary functions of biological colouration. While current research on biological colouration focuses mainly on organisms, understanding their ecological context requires quantifying habitat background – the visual stage on which ecology plays out. Animal perception of colour depends on background contrast, hue, and brightness, which affect clarity and can induce colour shifts. Environmental change altering habitat background colour may heighten risks such as predation, disrupted sexual selection, or intensified competition. Coral reefs are an ideal model system because degradation shifts habitats from diverse, colourful coral to relatively uniform algae. Coral reefs are increasingly mapped in 3D via photogrammetry, creating a growing repository of models for analysis (e.g., Operation Wallacea, MARS project) and facilitating new model generation. 3D models overcome the limitations of standard photographs, which often fail to capture the structural complexity and dynamic colouration of reef habitats. Reef fishes offer a rich testbed for investigating how shifts in habitat colour affect evolved colouration. Coral reefs also provide essential ecosystem services for billions yet are among the most threatened globally, with stressors like bleaching altering habitat colour and reducing background diversity.

References and Further Reading

1. Hemingson et al (2024) Analysing biological colour patterns from digital images: An introduction to the current toolbox. Ecology & Evolution (click here)
2. Koneru & Caro (2022) Animal coloration in the Anthropocene. Frontiers in Ecology & Evolution (click here)
3. Cuthill et al (2017) The biology of color. Science (click here)
4. Cooney et al (2022) Latitudinal gradients in avian colourfulness. Nature Ecology & Evolution (click here)
5. Caves et al (2024) Backgrounds and the evolution of visual signals. Trends in Ecology & Evolution (click here)

Overview and Background

In the last decade, Scotland’s rainfall increased by 9% annually and 19% in winter, with more water from extreme events, posing risks to the environment, infrastructure, health, and industry [Sniffer, 2021]. Urgent issues such as flooding, mass wasting, and water quality are closely tied to rainfall extremes [Sniffer, 2021]. Reliable predictions of extremes are, therefore, critical for risk management. Prediction of extremes, which is one of the main focuses of extreme value theory [Friederichs, 2010], is still considered one of the grand challenges by the World Climate Research Programme [Alexander et al., 2016]. This project will address this challenge by developing novel statistical, computationally efficient models that are able to predict rainfall extremes from the output of GPU-optimised climate models.

Methodology and Objectives

General Circulation Models (GCMs) are the primary tools for predicting future climate change [IPCC, 2023; and references therein]. While these GCM simulations are suitable for studies investigating climate dynamics and changes on coarse spatiotemporal scales, their skill in predicting local-scale climate and extremes remains very limited [IPCC, 2023]. Statistical Downscaling (SD) addresses this problem by linking coarse climate information to local-scale observational data [e.g., Hewitson et al., 2014; Mutz et al., 2021] using statistical models, which enables us to ”translate” GCM output to predictions that are more relevant for regional impact studies and adaptation measures. This project will leverage recent advances in SD for extremes [Cuba et al., 2024+] to develop a set of algorithms for predicting rainfall extremes in Scotland from the output of the latest GPU-optimised GCM ICON [Giorgetta et al. 2018]. These will be integrated into the user-focused, open-source tool pyESD [Boateng and Mutz, 2023]. Both teaser projects will rely on 2 datasets: 1) meteorological observations that capture rainfall extremes in Scotland (i.e., the ”predictand” dataset), and 2) a dataset used for SD model fitting (i.e., the ”predictor” dataset).

Teaser Project 1: “Perfect Prognosis” Approach

In the perfect prognosis approach, SD models are constructed from observation-based datasets for both the predictand and predictors. These models, therefore, capture real-world conditions and relationships, aiding model validation and improving our physical understanding of local-scale predictand variability. Another strength of this approach is the ability to couple the SD models to any GCM or dataset (e.g., Ramon et al., 2021), making them highly transferable. The predictor dataset for Teaser Project 1 will be ERA5 reanalysis data [Hersbach et al., 2020]. The SD models will then be coupled to 21st century simulations conducted with the GCM ICON [Giorgetta et al. 2018] to predict future changes in rainfall extremes in Scotland.

Teaser Project 2: “Model Output Statistics” Approach

Model Output Statistics also uses an observation-based predictand dataset, but the predictor dataset is simulated with climate models (e.g., GCMs). The relationships captured in the resulting SD models do not reflect physical processes as much as in the perfect prognosis approach, and the SD models are fine-tuned to a specific climate model. However, when used in tandem with this climate model, the approach often produces more accurate results and v. 2024.11.24 excels at climate model bias correction (e.g., Sachindra et al., 2014). The predictor dataset for Teaser Project 2 will be ICON simulations for the present-day climate. The SD models will then be coupled to 21st century ICON simulations to predict future changes in rainfall extremes in Scotland.

References and Further Reading

1. Alexander, L.V., Zhang, X., Hegerl, G. & Seneviratne, S.I. (2016). Implementation Plan for WCRP Grand Challenge on Understanding and Predicting Weather and Climate Extremes – the “Extremes Grand Challenge”. Version, June 2016 (click here)
2. Boateng, D. & Mutz, S. G. (2023). pyESDv1.0.1: an open-source Python framework for empirical-statistical downscaling of climate information. Geosci. Model Dev., 16, 6479–6514 (click here)
3. Coles, S. G. (2001). An Introduction to the Statistical Modeling of Extreme Values. London: Springer. Cuba, M.D., Wilkie, C., Scott, M. & Castro-Camilo, D. (2024+). Data fusion for threshold exceedances using a censored Bayesian hierarchical model. To appear
4. Friederichs, P. (2010). Statistical downscaling of extreme precipitation events using extreme value theory. Extremes, 13, 109-132
5. Giorgetta, M. A., Brokopf, R., Crueger, T., Esch, M., Fiedler, S., Helmert, J., et al. (2018). ICON-A, the atmosphere component of the ICON Earth system model: I. Model description. Journal of Advances in Modeling Earth Systems, 10, 1613–1637 (click here)
6. Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., et al. (2020). The ERA5 global reanalysis. Q. J. Roy. Meteor. Soc., 146, 1999–2049 (click here)
7. Hewitson, B. C., Daron, J., Crane, R. G., Zermoglio, M. F. & Jack, C. (2014). Interrogating empirical-statistical downscaling. Clim. Change, 122, 539–554 (click here)
8. IPCC (2023). Geneva, Switzerland, 35-115 (click here)
9. Mutz, S. G., Scherrer, S., Muceniece, I. & Ehlers, T. A. (2021). Twenty-first century regional temperature response in Chile based on empirical-statistical downscaling. Clim. Dynam., 56, 2881–2894 (click here)
10. Ramon, J., Lledó, L., Bretonnière, P.-A., Samsó, M. & Doblas-Reyes, F. J. (2021). A perfect prognosis downscaling methodology for seasonal prediction of local-scale wind speeds. Environ. Res. Lett., 16, 054010 (click here)
11. Sachindra, D. A., Huang, F., Barton, A. & Perera, B. J. C. (2014). Statistical downscaling of general circulation model outputs to precipitation – part 2: bias-correction and future projections. Int. J. Climatol., 34, 3282–3303 (click here)
12. Sniffer (2021): ‘Third UK Climate Change Risk Assessment Technical Report: Summary for Scotland’ (click here)

Overview and Background

Global climate and environmental change are increasingly resulting in weather extremes that impact society and infrastructure.  These extremes include stormier climates with increased wind speeds, precipitation events or drought, and temperatures (amongst other things). A team of University of Glasgow researchers are developing an Earth systems digital twin for exascale computing that works on GPU computers and uses weather forecasts to predict the cascading effect of climate change events on environmental systems. Our goal is to provide predictions, at the national or large scale, of the impacts of environmental extremes on natural and urban settings. This project is one, stand alone, component of this larger scale project.

In this project, you will focus on the glacier-bedrock interface. Mountain glaciers are changing rapidly worldwide in response to climate change. Glacier changes affect global trends in freshwater availability, contribute to recent sea level changes, and affect regional water resources over the twenty-first century. However, uncertainties remain in projecting such impacts in future climate change scenarios. A major source of these uncertainties is the lack of understanding of glacier sliding – the relative motion between glacial ice and underlying rocks (Zoet, L. K., & Iverson, N. R., 2020). In this project, you will develop a GPU-accelerated multiscale modelling framework for glacier sliding to tackle this problem.

Your job while working on this project will involve software development for simulating the relevant multiphysical processes, applying the model to historic data for validation, working in a team/workgroup environment, attending regular research group seminars, integrating diverse environmental and satellite data into your software, and learning new techniques through ExaGEO training workshops.

Methodology and Objectives

Subglacial system consists of ice, water, and rocks. Among these components, different processes and feedbacks operate at different spatial and temporal scales, making it a challenging computational problem to simulate. To tackle this problem, this project will adopt a multiscale modelling approach combining microscale modelling of the ice-bedrock interface with macroscale simulation of glacier dynamics and subglacial hydrology. A particular focus will be developing models for GPU architectures to enable high-resolution and scalable simulation.

Methods used in this project will involve numerical simulation using the open-source parallel finite element library MoFEM developed and supported at the University of Glasgow (Kaczmarczyk, Ł., et al, 2020).

Teaser Project 1:

This teaser project, conducted during the first year, will focus on developing a microscale model of the contact interface between glacier ice and bedrock. When the ice is compressed by its own weight against the bedrock, the roughness of both the ice and bedrock surfaces creates a complex contact problem where only isolated regions of the interface are in actual contact. Simultaneously, the remaining areas form free volumes that can be occupied by flowing or stagnant (trapped) water mixed with sediments. The sub-project will build upon a previously developed finite-element framework (Shvarts, A.G. et al., 2021) by enabling its application in a distributed-memory parallel computing environment and providing further GPU acceleration using the functionality available in the MoFEM library. Additionally, the framework will be enhanced to incorporate friction between the ice and bedrock. Using available data to calibrate the model, the extended framework will predict the shear strength of the interface as a function of various parameters, including the weight of the ice, surface roughness, sediment density, and water pressure. These predictions will be compared with existing phenomenological models of glacial sliding to refine and improve the latter.

Teaser Project 2:

This sub-project, also conducted during the first year, addresses the macroscale problem and focuses on the behaviour of the glacier as a whole. It will leverage the finite-element model implemented in MoFEM, constructed using available topological and geological data. The model will incorporate the results from the microscale simulations of the first sub-project, which map ice properties to interfacial shear strength, to accurately inform the macroscopic interface behaviour. Utilizing parallel computing with GPU acceleration, this sub-project will simulate large-scale glacier dynamics under varying environmental conditions, including changes in ice thickness, surface temperature, and basal water pressure. These simulations will provide critical insights into the glacier’s flow patterns, deformation, and sliding behaviour, enabling predictions of its response to climate change scenarios. The outcomes will also help validate and refine existing phenomenological models, improving their applicability to real-world glacier systems.

References and Further Reading

1. Zoet, L. K., & Iverson, N. R. (2020). A slip law for glaciers on deformable beds. Science, 368(6486), 76–78 (click here)
2. Kaczmarczyk, Ł., et al, 2020. MoFEM: An open source, parallel finite element library. The Journal of Open Source Software, 5(45) (click here)
3. Shvarts, A.G., Vignollet, J. and Yastrebov, V.A., 2021. Computational framework for monolithic coupling for thin fluid flow in contact interfaces. Computer Methods in Applied Mechanics and Engineering, 379, p.113738 (click here)

Overview and Background

Citizen science has enabled researchers to have access to an unprecedented large-scale, affordable, rich, and diverse data for researchers. However, many question the inherent biases and quality issues of community/citizen generated data. Within the field of ecology, communities such as iNaturalist and eBird collate hundreds of millions of georeferenced species observations from users around the world.

In order to address the challenges of quantifying the quality (accuracy, completeness, representation) of these data sources, this PhD uses state-of-the-art modelling and data science techniques, along with other data sources, such as high resolution remotely sensed raster data, to build a foundational understanding of data quality within volunteered ecological data. To achieve this, advanced techniques in computer vision, animal behaviour modelling and geo-AI will be employed, leveraging the considerable computational resources available to the ExaGEO project.

Methodology and Objectives

The two teaser projects will focus on a single aspect of data quality within crowd-sourced and community generated ecological data: identification of repeat observations. Datasets such as iNatrualist, Movebank, and eBird provide some indication of the spatial distribution of a wide variety of animal species, with some steps taken to ensure reasonable data quality. However, there is currently no protocol for identifying multiple observations of the same individual within a species.

The two teaser projects outlined below take different and complimentary approaches to estimating the likelihood that two observations of animals from the same species are in fact of the same individual. The output of these models may be used in downstream tasks later in the PhD project to assess overall data quality of the datasets and enable quantifiably measure the reliability of the outputs for biodiversity.

Methods Used:

1. Advanced computer vision (e.g. vision transformer models)
2. Spatially explicit AI models (e.g. Graph Attention Networks)
3. Animal behaviour modelling (e.g. ODEs/PDEs)
4. Remote sensed data processing/analysis- e.g. pixel or object-oriented classification of satellite data
5. Advanced statistical analysis (e.g. Markov chain Monte Carlo (MCMC), Hidden Markov, etc.)

Teaser Project 1

The first teaser project uses advanced computer vision techniques to identify similarities between photographs of animals and distinguish between individuals of the same species. To achieve this, a set of species specific keypoint models will be developed which are able to locate key identifiable features within an image (e.g. facial landmarks, joints, limbs). The relative location of these features, along with the geo-location of the observation and other data, will then be used within an unsupervised clustering model to identify likely observations of the same individual.

The success of this part of the project will rely on careful consideration of the taxon/a of study. Factors such as animal physiology, data availability and quality, and image processing and analysis techniques will play an important role in deciding this. Ultimately, the researcher will decide their area of study, after careful deliberation with the supervisory team. 

Teaser Project 2

The second teaser project involves the development of a spatio-temporal model for animal behaviour. A single observation of an individual animal consists of (at minimum) species information, a geolocation and a timestamp. By combining these with other data sources (e.g., land-cover data, remotely sensed data, estimated animal populations) and animal behaviour expertise, the researcher will construct a spatio-temporal model capable of estimating the likelihood that an observation of the same species at a different geo-location and time is the same individual animal.

The approach taken can use either mechanistic models (e.g. ODEs, PDEs) or statistical machine learning models (e.g. MCMC, Hidden Markov, GeoAI), or indeed may employ a combination of both (e.g. Particle/Kalman filtering, approximate Bayesian computation).

While these two projects have the same overall aim – namely, estimating the likelihood two observations are of the same individual – their methodologies are very different. Beyond the first year of the PhD, the models created in these projects may be combined, resulting in a single model which uses positional, temporal and visual information to identify repeat observations. Later work may use this model to investigate other aspects of data quality within volunteer ecological data in even more detail. This will be used for some of ongoing relevant project on balancing quantity-quality of crowdsourced data.

References and Further Reading

1. Lauer et. al. (2022), Multi-animal pose estimation, identification and tracking with DeepLabCut, Nature Methods
2. Hou et. al. (2020), Identification of animal individuals using deep learning: A case Study of giant panda, Biological Conservation
3. Vidal et. al. (2021), Perspectives on Individual Animal Identification from Biology and Computer Vision, Integrative and Comparative Biology
4. Wahltinez, O. and Wahltinez, S. J. (2024) An open-source general purpose machine learning framework for individual animal re-identification using few-shot learning, Methods in Ecology and Evolution
5. Laxton, M. R. et. al. (2022) Balancing structural complexity with ecological insight in Spatio-temporal species distribution models, Methods in Ecology and Evolution
6. Karppinen S. et. al. (2022) Identifying territories using presence-only citizen science data: An application to the Finnish wolf population, Ecological Modelling
7. Dorazio, R. M. and Karanth, K. U. (2017) A hierarchical model for estimating the spatial distribution and abundance of animals detected by continuous-time recorders, PLOS One
8. Supp et. al. (2021) Estimating the movements of terrestrial animal populations using broad-scale occurrence data, Movement Ecology
9. INaturalist
10. Movebank
11. EBird

Overview and Background

While only covering 3% of the Earth’s surface, peatlands store >30% of terrestrial carbon and play a vital ecological role. Peatlands are, however, highly sensitive to climate change and human pressures, and therefore understanding and restoring them is crucial for climate action. Multiscale mathematical models can represent the complex microstructures and interactions that control peatland dynamics but are limited by their computational demands. GPU and Exascale computing advances offer a timely opportunity to unlock the potential benefits of mathematically-led peatland modelling approaches. By scaling these complex models to run on new architectures or by directly incorporating mathematical constraints into GPU-based deep learning approaches, scalable computing will to deliver transformative insights into peatland dynamics and their restoration, supporting global climate efforts.

Teaser Project 1: Scalable Mathematical Modelling of Peatlands

Objectives: This project will explore how we can do scalable modelling and inference on mathematical models of peatlands. The project will take existing microscale models for peatlands and look at how we can perform mathematical optimisation to the learn the complex parameters of the mathematical model. The focus will then be on looking at how the model can be upscaled and improved, focusing on computational inference methods that will be applicable as the model gets expanded and becomes more computationally infeasible.

Methods: The project will use scalable mathematical processing and optimisation techniques, looking at how they compared to computational statistical inference methods such a Bayesian optimisation. The peatland model will be used for simulations and analysed to understand how it can be improved to model the complex non-linear processes. Future work as part of a potential PhD project would involve extending model and adapting it to be able to run in high performance computing environments and extending the optimisation techniques to work in this scenario.

PhD Project: The main purpose of the PhD project will be scaling up the peatland models, adding more features and scaling them to be able to run across computer clusters and on GPUs with the eventual aim of extending this to Exascale computing. Advanced mathematical techniques will be used to upscale from micro- to macroscale models incorporate nonlinear instabilities such as wrinkling and surface patterning. Computational methods will then be extended to focus on predicting long-term peatland behaviour under restoration scenarios and climatic stressors. The integration of experimental data for validation and refinement of models will ensure practical applicability.

Teaser Project 2: Mathematically Informed Machine Learning for Scalable Peatland Modelling

Objectives: This project will explore how we can used emulation techniques for scalable parameter inference and uncertainty quantification in an existing model for peatlands. We will use parallelised computing to run multiple simulations of the peatland model and then use GPU based deep learning methods to build an emulator. The emulator will provide a computational cheaper version of the original model, allowing us to use Bayesian inference in previously computational infeasible scenarios giving the ability to estimate the model parameters and their associated statistical uncertainty.

Methods: The project will make use of deep learning architectures that are designed to be specifically scalable to GPUs and eventually Exascale type infrastructures. Specifically, the emulation methods will use and compare deep neural networks and deep Gaussian processes to link model parameters to observed model outcomes. Optimisation and Bayesian inference will then be carried out using the emulator within the context of an inverse problem. Future work as part of a potential PhD project would involve extending these methods into more complex deep learning frameworks, e.g. physics informed machine learning or graph neural networks.

PhD Project: A potential follow-on PhD project would focus on incorporating the mathematical models directly into the deep learning structures and linking the model to real data. This could be achieved by making the models more scalable by replacing the mathematical finite element methods via GPU trained deep learning alternatives or through methods from the physics informed machine learning literature. . Linking the model to real data will also be computationally challenging, building either directly on the emulation methods from the initial project or through the mathematical informed machine learning methods that have been developed. Essentially, this project will aim to link this model to real world applications that can help us gain a better understanding of the structure of peatlands.

References and Further Reading

1. MathPeat Network
2. Peatland Factsheet
3. Effective Governing Equations for Poroelastic Growing Media
4. MPeat2D Peatland Model
5. Statistical finite element method
6. Neural Network Aided Multiscale Modelling
7. Surrogates (Introduction to Emulation)
8. Physics-Informed Machine Learning
9. Meshed Based Deep Learning

Overview and Background

Recent climate change has driven glacier retreat worldwide, releasing increasing volumes of sediments and meltwater into proglacial environments (Zhang et al., 2022). Combined with an increase in extreme weather events, this has triggered rapid geomorphic changes in proglacial fluvial systems (Heckmann et al. 2016). These changes pose significant risks to downstream areas, threatening infrastructure, food security, and ecological stability. Despite their importance, our understanding of 1) how proglacial rivers respond to increased meltwater and sediment influx from retreating glaciers and 2) how glacier-river system collectively responds to long-term climate trends and short-term weather extremes remains limited.

This PhD project seeks to address these knowledge gaps through advanced GPU-based numerical simulations. By developing and coupling models for sediment dynamics and glacier evolution, the research will explore the interplay between glaciers and proglacial rivers under varying climatic and environmental conditions. The student will focus on developing and validating the numerical model, designing and conducting simulations to understand key interactions and mechanisms, and applying the model in selected field locations to provide predictions.

Methodology and Objectives

This project aims to use GPU-based numerical simulations to investigate the responses of proglacial fluvial systems to both long-term climate changes and short-term weather variations. The work will involve developing new, efficient code for simulating sediment dynamics in rivers on GPU devices and/or coupling existing sediment dynamics models with a GPU-based glacial landscape evolution model. The simulations will be validated against field observations and integrated with environmental datasets to provide robust predictions of how proglacial river systems may evolve under future climate scenarios.

Teaser Project 1: Investigate sediment dynamics and geomorphic changes in proglacial fluvial system

As glaciers continue to retreat due to climate change, their downstream river systems experience dynamic and complex adjustments in sediment transport, channel morphology, and erosional/depositional patterns. The first teaser project aims to simulate and understand the rapid geomorphic changes in proglacial fluvial systems driven by variations in upstream meltwater and sediment fluxes.

The student will build upon existing sediment dynamics models such as SPACE (Shobe et al., 2017) and Eros (Davy et al., 2017) to create a GPU-based high-resolution 2D model tailored for proglacial rivers. This model will explicitly simulate sediment entrainment, transport, and deposition processes, allowing for detailed exploration of fluvial responses to varying upstream inputs. Advanced computational techniques, including parallel processing, will be employed to ensure efficiency and scalability.

Using this model the student will design and conduct scenario-based simulations to explore geomorphic changes in proglacial rivers under different conditions of meltwater and sediment input. A sensitivity analysis will then be performed to identify key parameters, such as slope gradients, channel geometry, and sediment grain size distribution, that influence sediment dynamics and channel evolution. Finally, the model will be calibrated and validated against field or experimental data to ensure accuracy and robustness, providing a foundation for broader application to various proglacial systems.

Teaser Project 2: Investigate the coupled evolution of glacier-river system driven by climate change and weather extremes

Recent climate change and increased weather extremes significantly impact both glacier dynamics and proglacial fluvial systems. By coupling glacier and sediment dynamics models, the second teaser project seeks to understand the coevolution of retreating glaciers and proglacial river systems under these influences, providing insights into their interconnected responses to climatic and extreme weather events.

The student will couple a GPU-accelerated glacier model (IGM; Jouvet and Cordonnier 2023) with a fluvial sediment dynamics model, such as SPACE or Eros. This integrated model will be able to simulate the response of the coupled glacier-river system to climate and weather variations.

The student will design and conduct a group of simulations covering a range of climate scenarios and investigate how climate-driven glacier retreat impacts proglacial river processes. The student will analyse the simulations to explore various scenarios of meltwater and sediment flux release under different climatic conditions, identifying distinct geomorphic responses of proglacial rivers and determining whether glacier retreat results in sedimentation, erosion, or a combination of both. Finally, the model outputs will be validated against field observations or experimental data, and sensitivity analyses will be conducted to identify the primary controls on river responses, providing robust insights into the coevolution of glaciers and proglacial river systems.

References and Further Reading

1. Davy, P., Croissant, T., & Lague, D. (2017). A precipiton method to calculate river hydrodynamics, with applications to flood prediction, landscape evolution models, and braiding instabilities. Journal of Geophysical Research: Earth Surface, 122(8), 1491–1512 (click here)
2. Heckmann, T., McColl, S., & Morche, D. (2016). Retreating ice: research in pro-glacial areas matters. Earth Surface Processes and Landforms, 41(2), 271–276 (click here)
3. Jouvet, G., & Cordonnier, G. (2023). Ice-flow model emulator based on physics-informed deep learning. Journal of Glaciology, 1–15 (click here)
4. Shobe, C. M., Tucker, G. E., & Barnhart, K. R. (2017). The SPACE 1.0 model: a Landlab component for 2-D calculation of sediment transport, bedrock erosion, and landscape evolution. Geoscientific Model Development, 10(12), 4577–4604 (click here)
5. Zhang, T., Li, D., East, A. E., Walling, D. E., Lane, S., Overeem, I., et al. (2022). Warming-driven erosion and sediment transport in cold regions. Nature Reviews Earth & Environment, 1–20 (click here)

Overview and Background

Scientific computing is crucial for understanding geophysical fluid flows, such as the geodynamo that sustains Earth’s magnetic field. This project will adapt an existing pseudo-spectral geodynamo code for magnetohydrodynamic simulations in rotating spherical geometries to GPU architectures, improving efficiency on modern computing systems and enabling simulations of more realistic regimes. This will advance our understanding of Earth’s geomagnetic field and its broader interactions, such as those with mantle heterogeneities. Evidence from seismology and geodynamics shows that the core-mantle boundary (CMB) is highly heterogeneous, influencing heat transport and geodynamo dynamics. By combining compressible, thermochemical convection with geodynamo simulations, this project will further investigate how deep slab properties affect the CMB heat flux, mantle heterogeneity, and the geodynamo.

Teaser Project 1: What is the impact of ancient slabs on core-mantle boundary heterogeneities and the geodynamo?

Evidence from seismology and geodynamics reveals that the lowermost mantle and the core-mantle boundary (CMB) are highly heterogeneous due to the presence of post-perovskite, large low shear wave velocity provinces and ancient, subducted slab material. CMB heterogeneity results in variable heat transport from the core and plays a key role in core and mantle dynamics, the geodynamo, and ultimately the Earth’s habitability. Previous work shows that the spatiotemporal evolution of the CMB heterogeneity is closely linked to deep slab dynamics (e.g., Heron et al., 2024, 2025), however these remain poorly understood. This teaser project will investigate the role of deep slab properties on temporal evolution of the deep mantle heterogeneity, the CMB heat flux and the geodynamo. This will involve modelling compressible, multiphase, thermochemical convection in a 3D spherical shell following the approach of Dannberg et al., (2024) and Heron et al., (2024, 2025) using the state of the art, open-source, adaptive mesh refinement, finite element software ASPECT (Heister et al., 2017). These models will include the subduction history over the last 1 billion year from Merdith et al., (2021) and will be supported by high resolution 3D regional models investigating the role of end-member slab properties (e.g., weak vs. strong slabs) on the CMB heterogeneity. Temporal variations in CMB heat flux from these models will then be analysed using spherical harmonics across the first 4 harmonic degrees similar to the approach of Dannberg et al., (2024) and used as thermal boundary condition for the geodynamo simulations. The goal is to expand teaser project 1 to investigate the influence the deep slab on core-mantle dynamics and the implications this has for magnetic field generation and the strength and frequency of polarity reversals.

Objectives:

1. Use global convection models to calculate the temporal evolution of heat flux at the CMB.
2. Investigate the influence of end member slab rheologies and geometries on the heat flux heterogeneity at the CMB.
3. Apply the calculated heat flux across the CMB from geodynamic models as a boundary condition to geodynamo simulations to investigate heterogeneity in magnetic field strength and the timing and frequency of magnetic field reversals.
4. Use GPU architecture to couple finite element mantle convection with geodynamo simulations.

Teaser Project 2: Spectral expansion transforms in spherical geometry

Modelling the geodynamo involves solving the coupled 3D, time-dependent, nonlinear Navier-Stokes equations, pre-Maxwell electrodynamics, and heat transfer equations for a rotating fluid. At present, the pseudo-spectral method is the most accurate and widely used numerical discretisation method in this context. The method requires applying physical to spectral space transforms which are generally in integral form and have been difficult to adapt to GPU architectures. With GPUs becoming increasingly powerful and accessible, this sub-project aims to port an existing versatile pseudo-spectral code for magnetohydrodynamic simulations in rotating spherical geometries to GPU systems.

Objectives:

1. Investigate alternative orthogonal polynomial basis function families that can be used to expand fields in spherical geometry, including Legendre, Jones-Worland, Jacobi and Galerkin.
2. Implement alternatives in and assess/compare convergence, stability and consistency of the resulting discretisations as well as their efficiency for GPU acceleration.

References and Further Reading

1. Dannberg, J., Gassmoeller, R., Thallner, D., LaCombe, F., & Sprain, C. (2023). Changes in core-mantle boundary heat flux patterns throughout the supercontinent cycle. arXiv preprint arXiv:2310.03229
2. Paul H Roberts and Eric M King. 2013. On the genesis of the Earth’s magnetism. Rep. Prog. Phys. 76 096801 (click here)
3. Gary A. Glatzmaier. 2014. Introduction to Modeling Convection in Planets and Stars: Magnetic Field, Density Stratification, Rotation. Princeton (click here)
4. Heister, T., Dannberg, J., Gassmöller, R., & Bangerth, W. (2017). High accuracy mantle convection simulation through modern numerical methods – II: Realistic models and problems. Geophysical Journal International, 210(2), 833–851 (click here)
5. Heron, P.J., Dannberg, J., Gassmöller, R., Shephard, G.E., & Pysklywec, R. N. (2025). The impact of Pangaean subducted oceans on mantle dynamics: passive piles and the positioning of deep mantle plumes. Gondwana Research
6. Heron, P.J., Gün, E., Shephard, G.E., Dannberg, J., Gassmöller, R., Martin, E., Sharif, A., Pysklywec, R. N., Nance, R.D., & Murphy, J.B. (2025). The role of subduction in the formation of Pangaean oceanic large igneous provinces. Geological Society London, Special Publications, 542(1)
7. Merdith, A.S. Williams. S.E., Brune, S., Collins, A.S., & Müller, D.R. (2021). Extending full-plate tectonic models into deep time: linking the Neoproterozoic and the Phanerozoic, Earth-Sci. Rev., 214 (click here)
8. Silva. L, Simitev, R., 2018. Pseudo-spectral code for numerical simulation of nonlinear thermo-compositional convection and dynamos in rotating spherical shells, zenodo.org, 1311203, 2018 (click here)

Overview and Background

Flood forecasting at regional and national scale is imperative to predicting the scale and distribution of floodwaters during extreme weather events, mitigating the impact on communities most at risk from flooding. The LISFLOOD family of surface water models have proved suitable to being parallelised at scale, allowing research and forecasting communities to take advantage of the previous generation of supercomputers, such as ARCHER. 

The increasing availability of high resolution topographic and meteorological data provides an opportunity to extend the capability of the LISFLOOD modelling framework to produce large-scale or high resolution flood forecasts at operational timescales – i.e producing model runs at sufficient lead-in times to alert communities to impending flood risk from forecasted extreme weather events. GPU-based exascale HPC systems provide the technological basis to develop forecast models delivering at operational timescales.

Methodology and Objectives

LISFLOOD is a family of hydrological models based on a 2D grid simulating rainfall-runoff. The water routing across a flood basin/river catchment is based on a simplified version of the shallow water (St Venant) equations. The model is process (physics) based, and there have been several implementations (see below), usually in C or C++, using a cellular automaton approach. These have been parallelised for CPU using OpenMP and in one spin-off project, MPI (see here for more details). 

The stencil-code library used in the previous CSE project, LibGeoDecomp, purports to have support for NVIDIA GPUs and CUDA (see here for more details). 

Teaser Project 1:

• Implement the hydrodynamic core of the LISFLOOD model on GPU hardware to demonstrate proof-of-concept that the current CPU parallelised code is portable to GPU hardware. 
• Methods for GPU parallelisation would include OpenMP offloading as initial approach to verify proof of concept. Project could then be extended to investigate CUDA bindings available in the libgeodecomp library.

Teaser Project 2:

• Profile, then optimise GPU ported code and test using case studies of UK extreme flood events, to indicate potential for near-realtime flood forecasting of GPU-enabled LISFLOOD code.
• This objective would aim to deliver a minimum-working example of a the GPU-ported flood model to deliver forecasts/re-analysis of a historic flood event in the UK within an operational timescale.

Development into a full PhD would involve further profiling and optimisation of the GPU code using either the libgeodecomp library, or another suitable GPU parallelisation framework. Delivering a proof-of-concept for a working flood forecast model at a regional scale would be a key aim of this project, demonstrating the potential to be used in operational flood forecasting systems. The full PhD may therefore look at workflow tools to integrate the various stages of forecast production such as: ingestion and pre-processing of data (i.e. from rainfall forecast/nowcasting data products), model scheduling on HPC systems, and post-processing of the outputs.

References and Further Reading

1. Coulthard, T.J., Neal, J.C., Bates, P.D., Ramirez, J., de Almeida, G.A. and Hancock, G.R., 2013. Integrating the LISFLOOD‐FP 2D hydrodynamic model with the CAESAR model: implications for modelling landscape evolution. Earth Surface Processes and Landforms, 38(15), pp.1897-1906
2. LISFLOOD model high level overview (click here)
3. Stencil Code for LibGeoDecomp (click here)
4. Open Source version of the C++ code developed by Declan Valters (click here)
5. Overview of an earlier project that developed an experimental version of the code for multi (CPU) node using stencil code (click here)

Overview and Background

Global climate and environmental change are increasingly resulting in weather extremes that impact society and infrastructure.  These extremes include stormier climates with increased wind speeds, precipitation events or drought, and temperatures (amongst other things). A team of University of Glasgow researchers are developing an Earth systems digital twin for exascale computing that works on GPU computers and uses weather forecasts to predict the cascading effect of climate change events on environmental systems. Our goal is to provide predictions, at the national or large scale, of the impacts of environmental extremes on natural and urban settings. This project is one, stand-alone, component of this larger-scale project.

In this project, you will develop and apply one component of the Earth system model.  We are seeking a student interested in GPU-accelerated large deformation modelling of landslides in soil-rock mixtures (SRM) along UK coasts. SRM is a naturally occurring material composed of high-strength rock fragments embedded in a matrix of low-strength soil. It is a common geological formation found in mountainous regions, river valleys and coasts. One example is the glacial till widely seen in the UK. SRM exhibits significant heterogeneity and anisotropy due to the random distribution and varying proportions of rock blocks and soil. In this project, we will develop a GPU-accelerated multifield plasticity simulation for modelling landslides in SRM.

Your job while working on this project will involve software development for simulating the relevant physical processes, applying the model to historic data for model evaluation, working in a team/workgroup environment, attending regular research group seminars, integrating diverse environmental and satellite data into your software, and learning new techniques through ExaGEO training workshops.

Methodology and Objectives

Methods used in this project involve multiscale modelling of SRM at the element level and large deformation modelling using multifield plasticity. The result from the multiscale modelling will be used to develop a constitutive model for SRM.

Teaser Project 1: Multiscale modelling of SRM

The mechanical behaviour of SRM is governed by the interaction between its components, with rock blocks contributing to structural stability and the soil matrix often controlling deformation and failure. Its unique characteristics, such as non-uniform strength, variable permeability, a wide range of particle sizes, and complex stress distribution, make SRM challenging to test and model. For instance, measuring the stress-strain relationship of SRM requires large equipment to accommodate the rock fragments in the testing cell. Developing such equipment is time-consuming and expensive. Therefore, we will use the multiscale approach to model the element response of SRM. At the mesoscale, elements of the SRM will be modelled to capture the detailed microstructure, including the distribution and properties of rock fragments and soil. This will be done using the finite element code MoFEM which is GPU compatible. The rock fragments will be modelled as non-deformable solids and soils will be modelled using a suitable elastoplastic model. The mechanical properties obtained from the microscale models will then be upscaled to the macroscale using homogenisation techniques. The multiscale modelling results will be validated and calibrated using experimental data to ensure accuracy. This involves comparing simulation results with laboratory tests reported in the literature. These simulations provide effective material properties that can be used in developing constitutive models for the SRM that are needed in large-scale modelling.

Teaser Project 2: Large deformation modelling of landslides using multifield plasticity

 The multifield plasticity developed at the Glasgow Computation Engineering Centre (GCEC) is a numerical method suitable for modelling large deformation problems, which eliminates the need for local integration of the elastoplastic model and can effectively exploit the computation power of GPUs. In the multifield framework, the balance of linear momentum, the flow rule, and the Karush–Kuhn–Tucker (KKT) constraints are formulated together within a variational framework. Beyond deformation, the plastic strain and the consistency parameter are treated as global degrees of freedom in the spatially discretised problem. To manage the increased number of global degrees of freedom, the method leverages the block sparse structure of the algebraic system and employs a customised block matrix solver designed to take advantage of modern hardware architectures. A constitutive model for the SRM will be implemented following the multifield plasticity and then used to model landslides in MoFEM. We will collaborate with research teams working on field observations and large-scale modelling in this development.

References and Further Reading

1. Lewandowski, K., Barbera, D., Blackwell, P., Roohi, A. H., Athanasiadis, I., McBride, A., … & Kaczmarczyk, Ł. (2023). Multifield finite strain plasticity: Theory and numerics. Computer Methods in Applied Mechanics and Engineering, 414, 116101
2. Gao, W. W., Gao, W., Hu, R. L., Xu, P. F., & Xia, J. G. (2018). Microtremor survey and stability analysis of a soil-rock mixture landslide: a case study in Baidian town, China. Landslides, 15, 1951-1961
3. Gao, W., Yang, H., & Hu, R. (2022). Soil–rock mixture slope stability analysis by microtremor survey and discrete element method. Bulletin of Engineering Geology and the Environment, 81(3), 121
4. Qiu, Z., Liu, Y., Tang, S., Meng, Q., Wang, J., Li, X., & Jiang, X. (2024). Effects of rock content and spatial distribution on the stability of soil rock mixture embankments. Scientific Reports, 14(1), 29088
5. Li, J., Wang, B., Wang, D., Zhang, P., & Vardon, P. J. (2023). A coupled MPM-DEM method for modelling soil-rock mixtures. Computers and Geotechnics, 160, 105508

Overview and Background

While heatwaves (sustained periods of hot weather) are a well-recognized public health hazard, growing evidence highlights an emerging risk from the co-occurrence of extreme temperature and air pollution[1,2]. The 2022 European heatwave, when the UK recorded its fist ever temperature >40°C, was accompanied by a widespread deterioration in air quality, with surface levels of ozone and other air pollutants exceeding safe limits across much of the continent[3]. The causal relationship between extreme temperature and extreme air pollution levels is complex, involving synoptic weather patterns affecting air movement, atmospheric chemistry, and pollutant emissions (e.g. from wildfires)[4,5]. In combination, these factors are not adequately understood or quantified but are important to detangle as the frequency and intensity of summer heatwaves will become more common due to climate change, meaning this ‘climate penalty’ for air quality could worsen[6]. This project will provide powerful new insight into the drivers of European extreme air pollution events during heatwaves and the associated health effects. This will be achieved by combining ultra-high-resolution model simulations of air pollutant behaviour, supported by satellite observations and other big observational datasets.

The successful candidate will join LEC’s vibrant atmospheric science research group: AtMOS

Methodology and Objectives

The two teaser projects are linked by an overarching theme (extreme air pollution during heatwaves), though are distinct in focus and employ different numerical modelling approaches. Teaser #1 involves a European-scale assessment with emphasis on improving scientific understanding of the underpinning processes responsible for elevating ozone during heatwaves. Teaser #2 provides a UK-scale assessment with emphasis on forecasting of extreme events and assessment of health effects. Both teasers will equip the student with key transferable skills around the acquisition/manipulation of atmospheric measurement and model datasets and the application of policy-relevant metrics to assess model performance.

Teaser Project 1:

Focussing on recent summer heatwaves, a Europe-wide assessment of the drivers of extreme ozone events will be performed. During the ‘teaser’, the temperature-ozone relationship will first be quantified by analysing measurement records from a large number of European monitoring sites, including the newly-available, extensive TOAR-II database of surface ozone observations (Year 1). The project’s principal modelling tool will be the FRSGC/UCI chemical transport model (CTM) that is developed and maintained in Lancaster. During the teaser, the ability of the model to capture elevated summertime ozone will be examined using the 2022 heatwave as a case study (Year 1).

If developed into a full project, the work will be expanded in scope to cover other notable European heatwaves (e.g. summer 2018). In addition to surface measures, the behaviour of ozone and the model’s ability of to capture it, will be further evaluated with satellite measurements of atmospheric composition (e.g. ozone from the GOME-2 and IASI instruments). The focus of the project in Years 2/3 will be to detangle the drivers of elevated ozone using carefully designed model sensitivity experiments. These will allow, for example, the significance of temperature-induced increases in ozone precursors emissions to be explored, including biogenic volatile organic compound emissions from vegetation (e.g. isoprene) and wildfires (CO, NOx), along with assessment of the long-range transport of ozone from outside of continental Europe (including from the stratosphere). 

Objectives:

1. Characterize the European ozone-heatwave response across multiple summers using a suite of surface and satellite measurements.
2. Assess the ability of the FRSGC/UCI atmospheric chemical transport model to capture extreme ozone events and the observed ozone-temperature relationship.
3. Interpret the observed ozone-heatwave response using high resolution model simulations. Explore multiple factors, including the relative role of meteorological versus chemical drivers, the effects of model horizontal resolution, and other assumptions.

Teaser Project 2:

Process-based air quality models are frequently used to ‘hindcast’ the state of air quality over a given region, providing information required to assess the impact of changing air pollutant levels on public exposure and health. Additionally, such models are now increasingly used to alert the public and health care providers in advance of upcoming air pollution episodes (i.e. ‘forecast’). Like weather forecasts, air quality forecasts may be provided up to several days ahead, with forecast confidence generally decreasing with increasing lead time. As forecast skill is often inadequate, particularly for the most ‘extreme’ episodes, a body of literature on possible approaches to ‘bias correct’ forecasts (before they are issued) has emerged, some of which involve near real-time data assimilation[7-9].

This project will examine the ability of WRF-Chem to simulate UK air quality in both hindcast and forecast modes, with an emphasis on heatwave periods. WRF-Chem is a well-evaluated and widely adopted model suitable for high resolution simulations at the country scale. During the project’s ‘teaser’ part (Year 1), the skill of WRF-Chem to forecast surface ozone will be assessed considering a range of forecast lead times (24 to 96 hours) and by applying a range of key metrics (e.g., hit rate, false alarm rate etc.). For model evaluation, we will utilise the UK’s extensive ‘AURN’ network of air pollutant measurements. If developed into a full project, the student will explore the effectiveness of a range of bias correction techniques (Year 2), with emphasis on developing and implementing a scheme that improves the tail of the ozone distribution in both hindcast and forecast set ups. Bias-corrected hindcasts will be produced and the annual mortality burden attributable to long-term air pollutant exposure will be quantified[10] (Year 3). This analysis will be performed in time slices from ~1990 to present, allowing the effectiveness of air quality legislation on health burdens over time to be quantified.

Objectives:

1. Evaluate the skill of the WRF-Chem model in reproducing surface ozone and other air pollutants over the UK during recent heatwaves.
2. Assess the efficacy of a range of bias correction techniques (e.g. ‘quantile mapping’) applied to WRF-Chem hindcasts and forecasts.
3. Produce bias-corrected hindcasts of key air pollutants and quantify the associated human health effects of air pollution in the UK over time.

References 

1. Schnell, J.L., and Prather, M.J. (2017). Co-occurrence of extremes in surface ozone, particulate matter, and temperature over eastern North America. Proc. Natl. Acad. Sci., 114, 2854-2859 (click here)
2. Gouldsbrough, L., Hossaini, R., Eastoe, E., & Young, P.J.Y. (2022). A temperature-dependent extreme value analysis of UK surface ozone, 1980-2019. Atmos. Env., 273, 118975
3. Copernicus scientists warn of very high ozone pollution as heatwave continues across Europe
4. Pope, R. J., et al. (2023). Investigation of the summer 2018 European ozone air pollution episodes using novel satellite data and modelling, Atmos. Chem. Phys., 23, 13235-13253 (click here)
5. Otero, N., Jurado, O. E., Butler, T., and Rust, H. W. (2022). The impact of atmospheric blocking on the compounding effect of ozone pollution and temperature: a copula-based approach, Atmos. Chem. Phys., 22, 1905-1919 (click here)
6. Doherty, R.M., Heal, M.R., and O’Connor, F.M. Climate change impacts on human health over Europe through its effect on air quality, Environ. Health, 16 (click here)
7. Neal, L.S., Agnew, P., Moseley, S., Ordóñez, C., Savage, N.H. and Tilbee, M. (2014). Application of a statistical post-processing technique to a gridded, operational, air quality forecast, Atmos. Env., 98, 385-393 (click here)
8. Staehle, C., et al. (2024). Technical note: An assessment of the performance of statistical bias correction techniques for global chemistry–climate model surface ozone fields, Atmos. Chem. Phys., 24, 5953-5969 (click here)
9. Gouldsbrough, L., Hossaini, R., Eastoe, E., Young, P.J.Y. & Vieno, M. (2023). A machine learning approach to downscale EMEP4UK: analysis of UK ozone variability and trends. Atmos. Chem. Phys., 24,  163-3196 (click here)
10. Macintyre, H.L., et al. (2023). Impacts of emissions policies on future UK mortality burdens associated with air pollution. Environ. Int., 174, 107862 (click here)

Further Reading

1. How UK’s record heatwave affected air pollution
2. World meteorologists point to ‘vicious cycle’ of heatwaves and air pollution
3. Schnell, J.L., and Prather, M.J. (2017). Co-occurrence of extremes in surface ozone, particulate matter, and temperature over eastern North America. Proc. Natl. Acad. Sci., 114, 2854-2859 (click here)
4. Pope, R. J., et al. (2023). Investigation of the summer 2018 European ozone air pollution episodes using novel satellite data and modelling, Atmos. Chem. Phys., 23, 13235-13253 (click here)

Overview and Background

This PhD studentship focuses on developing GPU-accelerated models of magmatic processes that underpin volcanic hazards and magmatic resource formation. These processes span sub-millimeter mineral-melt-fluid interactions up to kilometer-scale magma dynamics and crustal deformation. Magma is a multi-phase mixture of solids, silicate melts, and volatile-rich or other fluids, interacting in complex thermo-chemical-mechanical ways. The project will contribute one component of a hierarchical, multi-scale modelling framework using advanced GPU-based techniques. In this project, you will focus on developing a system-scale model of fluid exsolution and extraction from a crystallising magma body with implications for volcanic unrest preceding eruptions and the genesis of magmatic-hydrothermal deposits of critical metal ore.

Your work will include software development, integrating and interpreting field and experimental data sets, attending regular seminars, collaborating within a research team, and receiving training through ExaGEO workshops.

Volcanic eruptions originate from shallow crustal magma reservoirs built up over long periods. As magma cools and crystallises, it releases fluid phases—aqueous, briny, or containing carbonates, metal oxides, or sulfides—whose low viscosity and pronounced density contrasts drive fluid segregation. This fluid migration can trigger volcanic unrest or concentrate metals into economically valuable deposits. The micro- to meso-scale distribution of fluids—discrete droplets versus interconnected drainage networks—crucially depends on crystal and melt properties. Direct observations are challenging, so high-resolution, GPU-accelerated simulations provide a way to understand these complex and dynamic systems.

Methodology and Objectives

Modelling volcanic systems is challenging due to the multi-scale nature of their underlying physical and chemical processes. System-scale dynamics (100 m to 100 km) emerge from interactions involving crystals, melt films, and fluid droplets or channels on micro- to centimetre scales. To link these scales, this project uses a hierarchical approach: (i) direct numerical simulations of granular-scale phase interactions, (ii) deep learning-based computational homogenisation to extract effective constitutive relations, and (iii) system-scale mixture continuum models applying these relations to problems. All components leverage GPU-accelerated computing and deep learning to handle direct simulations at local scales, train effective constitutive models, and achieve sufficient resolution at the system scale.

In this project the candidate will develop and apply a novel system-scale three-phase flow model informed by effective constitutive models derived from granular-scale simulations and computational homogenisation. The model will build on a recent multi-phase reaction-transport theory framework [1] and numerical treatment [2] which will be implemented in a GPU-accelerated algorithm built on cutting edge Julia packages [3]. To inform system-scale reaction and transport rates the simulations will utilise constitutive models derived from granular-scale direct simulations and computational homogenisation (delivered by partner projects). The simulations will be used to systematically investigate the role of a regime transition in the transport of fluid phases through subvolcanic magma bodies from disconnected bubble migration to interconnected channelised drainage [4].

Within this framework, the student will start by working on two “teaser” projects to gain familiarity with different techniques and data, then choose how to further develop and focus their research.

Teaser Project 1:

This teaser project, conducted over the first year, will focus on the implementation of a mechanical three-phase (solid+liquid+liquid/vapour) transport model using GPU-accelerated algorithms in Julia. The implementation will follow from previous work demonstrated in a serial Matlab prototype [2] and use a staggered-grid finite-difference method to discretise the set of underlying PDEs in combination with a matrix-free iterative solution approach demonstrated to be highly efficient when run on massively parallel GPU infrastructure. The model will be structured such that it can switch between using traditional constitutive models as analytical functions of system variables [1,2] and querying trained neural nets which output flux and transfer rates given gradients and phase deviations in system variables.

Teaser Project 2:

This teaser project, conducted over the first year, will focus on implementing and calibrating a multi-component petrological model to represent fractional crystallisation and fluid exsolution in a chosen volcanic context (to be determined). The petrological model will follow an approach utilising thermodynamics-inspired fitting functions [5] and will be calibrated against the output of an energy-minimising thermodynamic equilibrium solver [6] using machine learning tools. This established approach for formulating an approximate but robust and efficient form of phase equilibrium model will be compared to a novel approach of training a neural network to take pressure, temperature, and element composition of magmatic materials and returning the proportions and compositions of stable phase assemblages.

References and Further Reading

1. Keller, T. and Suckale, J., 2019. A continuum model of multi-phase reactive transport in igneous systems. Geophysical Journal International, 219(1), pp.185-222
2. Wong, Y.Q. and Keller, T., 2023. A unified numerical model for two-phase porous, mush and suspension flow dynamics in magmatic systems. Geophysical Journal International, 233(2), pp.769-795
3. ParallelStencil; ImplicitGlobalGrid
4. Degruyter, W., Parmigiani, A., Huber, C. and Bachmann, O., 2019. How do volatiles escape their shallow magmatic hearth?. Philosophical Transactions of the Royal Society A, 377(2139), p.20180017
5. Riel, N., Kaus, B.J., Green, E.C.R. and Berlie, N., 2022. MAGEMin, an efficient Gibbs energy minimizer: application to igneous systems. Geochemistry, Geophysics, Geosystems, 23(7), p.e2022GC010427