University of Southampton

Southampton is a research-intensive university and a founding member of the Russell Group, and home to 25,000 students and 6,000 staff.

The university ranked 8th in the UK for research intensity and 81st worldwide (QS, 2024), with continued substantial growth in research funding and 7th in the UK for the total value of The Engineering and Physical Sciences Research Council (EPSRC) grants. The university is also a founding partner of SETsquared, the university business incubator ranked 1st in the world (UBI Global).

Research is divided into four broad themes: Agriculture, Food and Health; Environment; Heritage and Creativity; and Prosperity and Resilience.

Ocean & Earth Science ranked 38th and Statistics & Operational Research ranked 48th in the world (QS, 2024).

The university is a three-times winner of the Queen’s Anniversary Prize,

Southampton holds a Silver Athena Swan award for our commitment to gender equality and continued efforts to support the career aspirations of females.

Potential PhD projects – 2027

1. Decadal changes in the North Atlantic Ocean marine environment unravelled with causal deep learning and dynamic network modelling
   The North Atlantic Ocean is strongly impacted by climate change, and data show ongoing changes in its nutrient distributions and marine ecosystem. This project aims to identify the drivers of these changes by applying deep learning techniques for causal attribution and network analysis to a large variety of oceanographic data.
   Partner: National Oceanography Centre (NOC)
2. Scalable Interior-Point Methods for Constrained Data Assimilation in Weather and Climate
   Weather and climate forecasts are initialised by solving enormous least square problems (data assimilation). This project develops scalable interior-point methods and novel preconditioners—classical and statistical / probabilistic—for the constrained least square problems that enforce physical bounds, producing more accurate, physically consistent analyses.
   Partner: National Oceanography Centre (NOC)
3. Do AI Weather Models Learn the Dynamics of Extreme Events?
   This project will test whether AI weather models learn the physical development of ex-treme events or simply reproduce patterns that appear in their training data. We will ex-plore how AI represents key stages of cyclones, heatwaves, blocking and atmospheric riv-ers, thus identifying when AI weather forecasts become unreliable.
4. Mapping oceanographic variables with data-driven and physics-informed statistical methods
   Ocean variables such as temperature, salinity, and nutrients vary across a wide range of spatio-temporal scales. Ocean measurements are often patchy, making estimation of underlying structures in space and time challenging. In this project, we aim to develop novel and improved statistical methods for the spatio-temporal mapping of these properties.
   Partner: National Oceanography Centre (NOC).
5. Machine learning-enhanced stochastic programming for long-term energy storage management under extreme weather and climate risk
   This project will develop machine learning-enhanced stochastic programming, decision-focused scenario generation and scalable solution algorithms for long-duration energy storage management under extreme weather and compound climate risks. The research will address risk-averse, multi-horizon optimisation of large-scale energy systems, enabling efficient and resilient storage investment and operational decisions.
6. Accurate and efficient modelling of cooling technologies for sustainable data-driven computing
   This PhD candidate will develop new, physically-accurate mathematical models to make computing—such as global AI computing in data centers—more energy efficient. You will apply analytical techniques such as asymptotic and complex variable methods, validated by numerical simulations, to help design and improve (multi-physics) thermal cooling technologies such as vapor chambers, helping to reduce the environmental impact of the large-scale computation boom.
7. Can We Distinguish Between Physical and Artificial Breakpoints in Climate Data Records?
   Climate records contain abrupt changes arising from both the climate system and the observing system. This project will use classical statistical approaches, AI, climate model simulations and historical observations to identify and interpret breakpoints, helping distinguish physical climate transitions from observational artefacts.
   Partner: National Oceanography Centre (NOC).
8. Detecting deep water formation in the North Atlantic using remote sensing, machine learning and mathematical methodsDeep convection has a key role in dense mass formation, and ultimately impacts the Atlantic Meridional Overturning Circulation. This project will investigate these events by developing a machine learning approach to integrate in situ measurements, remote sensing, and high-resolution ocean simulations.
   Partner: National Oceanography Centre (NOC).
9. Reconstructing High-Resolution Climate Records: Deconvolution of Marine Sediment Data to Unveil Past Climate Dynamics

   Palaeoclimate records from marine sediments are vital for understanding past climate changes but are often distorted during deposition and measurement processes. This project develops Bayesian and AI-based deconvolution methods, validated through experiments, to restore high-resolution and accurate climate reconstructions to confront and inform climate models for improved future predictions.
   Partner: National Oceanography Centre (NOC).

10. Can individual environmental responses predict associations between speciation, extinction and climate change?
    You will build integral projection models for species with complex demography, using stochastic processes to explore the evolutionary stable strategies that lead to species’ emergence and persistence. Your models will be compared against state-of-the-art integrated geochemical and computer tomographic data to make rigorous inferences of pivotal eco-evolutionary processes.
11. Laser Focused: recovering past climate from blurred laser-ablation images using statistical inversion and machine learning
    Geochemical imaging by laser ablation lets us read past climate information from coral skeletons, but the measurement process blurs the chemical signals we want to recover. In this project you will develop statistical inversion and machine-learning methods to deblur these images, revealing fine-scale climate information currently hidden from view.
    Partner: National Oceanography Centre (NOC) Liverpool
    
12. Tracing Cloud Formation Pathways Across Earth’s Atmospheric History
    

    This PhD project investigates cloud feedback and atmospheric composition in Earth’s climate history, focusing on the Archean aeon. Using atmospheric modelling, palaeoclimate analysis, and cloud physics, the student will develop new tools for 3D climate models, advancing understanding of climate change, atmospheric dynamics, and the role of clouds.

13. Sea-ice characterisation from autonomous underwater vehicles
    

    This PhD will develop innovative acoustic methods to characterise sea-ice, critical for climate, ecosystems, and safe navigation. The project combines polar science, acoustic data analysis, mathematical theory and modelling, providing unique training at the intersection of mathematics, oceanography and engineering in real-world Polar research.
    Partner: National Oceanography Centre (NOC), Defence Science and Technology Laboratory (Dstl) and Imperial College London (ICL)

14. Probabilistic Representation Discovery for Long-Tailed Visual Recognition
    

    You will develop probabilistic, label-efficient methods to discover structures in long-tailed environmental image data. You will explore how / your initial approach will leverage the learning of prototypes and hierarchies from unlabeled data, improves recognition of rare phenomena, and supports open-world detection across ecology and ocean observation for biodiversity monitoring.
    Partner: National Oceanography Centre (NOC)

15. Sea Ice decline interaction with ocean eddies and turbulence
    The project examines how sea ice decline in the warming climate changes energy input into ocean and ocean turbulence. The project makes use of observations, continuum mechanics theory, ultra-high-resolution models, and novel mathematical techniques as Artificial Intelligence and Machine Learning to quantify mechanisms which alter ocean turbulence at multiple scales.
    Partner: National Oceanography Centre (NOC)