ExaGEO PhD programme: advancing environmental science with Exascale computing

14 January 2025

By equipping researchers with cutting-edge computational skills, the ExaGeo PhD programme will advance scientific understanding and provide innovative solutions to global environmental issues.

Image shows the results of discrete element simulations using the MFIX code. Credit: Eric Breard, School of Geosciences, UoE.

The mission of the ExaGEO initiative is to harness the potential of Exascale computing to address global environmental change by training PhD students. The initiative will focus on developing expertise in advanced areas such as GPU-accelerated environmental modelling and large-scale Earth observation data analysis. ExaGEO will support not only Exascale computing model development and data analysis, but also multidisciplinary training of students in Earth system processes and applications.

Each student will be positioned within a supervisory team consisting of multidisciplinary supervisors: one computational, one domain expert, and one from an Earth or environmental, and/or social science research background. This ‘team-based’ supervisory approach is designed to enhance multidisciplinary training.

The University of Edinburgh, through EPCC, is the primary institution for four of these projects.

ExaGeo projects based at EPCC

GPU-accelerated high-fidelity hydrodynamics modelling for tidal energy resource and environmental impact assessment

EPCC supervisor: Joe O'Connor

Tidal energy offers a predictable and sustainable energy source, and this is driving increased interest in its development. However, as tidal energy deployments are scaled up, a greater burden is placed on the local environment/ecosystem. High-fidelity hydrodynamic modelling tools are essential for predicting and mitigating environmental impacts, while also maximising energy extraction, to ensure this limited resource is used in a responsible way. However, these models are computationally demanding and many existing tools were developed for traditional (CPU-based) HPC systems. With the advent of large GPU-based exascale machines, there is a need to prepare existing codes for this new HPC paradigm. 

This PhD will develop advanced computational techniques to improve and accelerate hydrodynamic modelling capability for tidal energy applications, with an emphasis on GPU-acceleration and exascale computing. As well as futureproofing existing models, this will unlock new types of simulations (eg multi-physics) and workflows (eg optimisation, uncertainty quantification, data assimilation) that are otherwise extremely challenging with today's methods. While based at EPCC, the successful candidate will also work closely with tidal energy experts in the School of Engineering.

Investigating the rheology of volcanic granular flows with GPU-based Discrete Element Method

EPCC supervisor: Kevin Stratford

The aim of this project is to investigate the properties of geophysical materials as they deform and flow (that is, their rheology). Systems of interest include pyroclastic flows produced during volcanic eruptions, landslides, and debris flow produced during flooding. All these represent significant natural hazards.

The project will use the discrete element method, a computational method that represents materials as aggregates of individual particles. This allows the complex interactions which determine the overall flow behaviour to be captured. As very many particles are needed to construct realistic systems, significant computational effort is the order of the day.

Developing large-scale hydrodynamic flood forecasting models for Exascale GPU systems

EPCC supervisor: Mark Bull

Numerical flood forecasting, at regional and national scales, to predict the scale and distribution of floodwaters during extreme weather events, is vital for mitigating the impact on communities most at risk from flooding. As with weather forecasting, flood forecasting is highly time-sensitive, so the performance of forecast models is very important for achieving fast turnaround in an operational context.

This PhD project will work on the LISFLOOD family of hydrological models, based on a 2D grid simulating rainfall-runoff. The initial aim will be to port the code to the latest GPU hardware, with further ambitions to improve the code performance and to optimise the whole forecast workflow, including data ingestion and pre-processing, as well as post-processing model output into human-readable forecast products.

Exploring solver approaches for climate and fluid simulations

EPCC supervisor: Adrian Jackson

Numerical solvers are the core of many computational simulation approaches. Weather and climate simulations are prime examples of these, with numerical/climate simulation packages such as the Met Office’s Unified Model (UM) containing various “dynamical core” implementations. Modern computer models of the atmosphere include many complex physical processes that each have local influences and feed back into the general circulation. At the heart of these models, however, is the solution of the dynamical equations of motion (Newton’s laws applied to a gas). For this reason, the model component that solves these equations is called the “dynamical core".

Whilst these solvers are the work of continual upgrades and improvements, they all follow similar approaches. However, alternative numerical approaches do exist, such as contour-based schemes, which move away from the grid based spatial discretisation and to potentially more functional implementations. In theory these should be significantly more efficient and deal with complex phenomenon such at turbulence in a more natural manner. However, they have yet to replace existing methods for simulation.

This project aims to investigate the computational benefits and drawbacks of these different types of numerical approaches, especially with respect to modern computing hardware (GPUs and other accelerators) that will make up future Exascale systems. We aim to both understand the underlying computational requirement of different numerical approaches, how they map to computational systems, and what other functionality is required to enable upgrading the numerical approaches of our model weather, climate, and environmental simulation systems. The promise is for a 100x+ efficiency improvement in computational approaches if this can be achieved.

Further information

Please note that applications to join the ExaGEO programme and receive PhD studentship funding must be submitted via the University of Glasgow’s online application system, regardless of your intended host institution.

Applications will close on 17th February 2025.

Full details are available from the ExaGeo website. To discuss a potential project, please contact the relevant EPCC supervisor.

Supported by funding from UK Research and Innovation’s Natural Environment Research Council (NERC), the programme brings together a consortium of three institutions — the University of Glasgow, the EPCC at the University of Edinburgh, and CEEDS at Lancaster University — alongside over 200 experts from academia and industry.

Image above shows the results of discrete element simulations using the MFIX code. Credit: Eric Breard, School of Geosciences, University of Edinburgh.