Application Deadline: 06 January 2027
Details
This PhD is part of the Net Zero Polar Science DTP, which aims to make polar science possible in a net zero world. For further details visit https://nzps-dtp.ac.uk/
Project Summary
Antarctica’s floating ice shelves are a key component of the climate system as they act to hold back (or buttress) the grounded ice of the Antarctic Ice Sheet. Ice shelf loss can lead to the grounded ice accelerating into the ocean and contributing to global sea level. Understanding when and why this happens is one of the biggest open questions in determining future sea level rise. Ice shelves have been known to catastrophically collapse, with melting on the surface having been suggested as a key trigger. Surface melting can lead to the formation of melt lakes and rivers. These both change the energy balance of the ice shelf and trigger fracture of the ice that can then lead to dramatic collapse. A physics-based numerical model, MONARCHS, has been developed to allow us to investigate these processes. However, due to the including of many detailed physical processes, the model is computationally expensive to run. This makes it harder for us to answer important science questions- which ice shelves are going to collapse and when?
This PhD sits at the intersection of artificial intelligence and polar physics. A proof-of-concept emulator of MONARCHS already exists: one that can recreate the skill of the MONARCHS model whilst running thousands of times faster, and thus at a fraction of the computational cost. Building on these steps will lead to insights into the future stability of Antarctica’s ice shelves, providing key next steps into understanding the processes that lead to sudden ice shelf collapse, and ultimately reducing the uncertainty in future sea level change. For example, is a hybrid physics-AI MONARCHS approach like Google’s NeuralGCM model, where a physics core is retained and uncertain parts of the model learned with AI, the best to combine the benefits of both world? Might XAI methods help understand what it has learned? What new processes about Antarctic ice shelves can we now understand given the vastly reduced cost of compute? How can satellite observations be brought in and used with AI to refine this new emulator?
A Net Zero case study will be carried out during the project to determine the impact the emulator has on reducing the computational (and thus carbon) cost of carrying out these modelling studies. This will allow the PhD researcher to show the direct, measurable sustainability impact of their work along side answering some of the key questions in polar science.
Research Objectives
- Determine the most computationally expensive processes in a model of ice shelf surface hydrology, MONARCHS, that are suitable for emulation, then develop AI emulators and hybrid physics-AI versions of MONARCHS (where a physics core is retained). Establish stability, physical conservation, and trust of the hybrid model through XAI methods or other approaches. Use the new MONARCHS emulator to determine current and future ice shelf states across Antarctica.
- Net Zero Case Study: The project will investigate the impact of emulation on a computationally heavy model, allowing science questions to be answered for less computational cost that is currently possible. The PhD researcher will track computational costs before and after the emulator is implemented, allowing the reduction in compute time to be quantified.
Technical Approaches
This project combines a physics-based ice shelf model with artificial intelligence. The student will first explore the physics of MONARCHS, judging which parts suit emulation and which should remain as physics. Building fast neural emulators, they will explore hybrid designs with learned components inside the physical model, test conservation of physical quantities, stay numerically stable and generalise. Further themes include quantifying uncertainty, probing what the networks have learned using existing or new explainable-AI theory, and incorporating observational data into the learned components, potentially after downscaling. The student will gain skills as they choose across polar physics, modelling and machine learning.
This project combines a physics-based ice shelf model with artificial intelligence. The student will first explore the physics of MONARCHS, judging which parts suit emulation and which should remain as physics. Building fast neural emulators, they will explore hybrid designs with learned components inside the physical model, test conservation of physical quantities, stay numerically stable and generalise. Further themes include quantifying uncertainty, probing what the networks have learned using existing or new explainable-AI theory, and incorporating observational data into the learned components, potentially after downscaling. The student will gain skills as they choose across polar physics, modelling and machine learning.
Training Opportunities
The PhD researcher will have the opportunity to spend time working with machine learning experts in Cambridge (based at both the British Antarctic Survey and The University of Cambridge). They will also be encouraged to apply to specific polar science related summer schools.
Academic Background
This project would be suitable for students with a degree in computer science, physics, mathematics or a closely related subject.
Eligibility
For entry to PhD study, applicants are expected to have at least one of the following:
• a first or upper second (2:1) class honours undergraduate degree in a relevant subject, or an equivalent international qualification,
• a relevant master’s qualification or equivalent evidence of prior professional practice.
International applicants and candidates from non-English speaking countries will need to meet the minimum language requirements for admission onto the programme of study for their Home institution.
How to Apply
Informal enquiries about the project and your application should be addressed to the project supervisor, Sammie Buzzard, sammie.buzzard@northumbria.ac.uk
After you have discussed your application with the project supervisor and read the NZPS application guidance, you should:
1) Complete the online NZPS Application Form by 10.00 (GMT) Wednesday 6th January 2027
2) Submit any additional application documents in the requested format to nzps@northumbria.ac.uk by the closing date.
If you require any additional assistance in submitting your application or have any queries about the application process, please don’t hesitate to contact us at nzps@northumbria.ac.uk
Funding Notes
Funding is available to Home/UK and international (including EU) students, subject to the successful completion of quality assurance checks and UK Visa and Immigration (UKVI) compliance requirements. This includes a full stipend at UKRI rates (for 2026/27 FT study this is £21,805 per year), full tuition fees and an annual Research Training and Support Grant (RTSG).
Studentships are also available for Home applicants who wish to study part-time in combination with work or personal responsibilities. Please note: additional costs may apply for international applicants, including visa costs and NHS surcharge.
