skip to content

AI for the study of Environmental Risks (AI4ER)

UKRI Centre for Doctoral Training

2021 Cohort (MRes)

Grace Beaney Colverd

PhD project title: Modelling the material and emissions impacts of the interventions required to decarbonise the building stock.

Supervisors: Dr Jonathan Cullen, Dr Ronita Bardhan

PhD Project description: We propose a novel assessment of the UK housing stock for retrofit using multi-task learning on a per-house basis. The goal here is to produce an analysis that is more detailed than top-down approaches but covers much more ground than traditional bottom-up approaches. The output will be a ranked list that can be provided at varying geographic levels, in order to help prioritise and plan retrofit in the UK.

Prior to joining the CDT, I worked in strategy consulting in London, as an Associate Managing Consultant at Mastercard, and before that as a Consultant with Applied Predictive Technologies. My focus was delivering data-driven analytics to Retail, Banking, and Governmental clients. I had a particular interest in financial inclusion and its impact on intersectional financial equality. 

I completed my undergraduate and masters degrees at The Queen’s College, University of Oxford, where I gained an MPhysPhil in Physics and Philosophy. My final thesis covered topics within string theory: properties of Calabi-Yau manifolds encoded within 2-interior point polytopes, and the results can be found published in the Journal of High Energy Physics. I am interested in practical applications of machine learning to protect earth systems, it’s applications to policymaking and popular science communication.

Hamish Campbell

PhD project title: An AI enabled peatland regeneration prioritisation map of the East Anglian Fens

Supervisors: Prof. David Coomes, Dr. Oscar Aldred, Prof. Srinivasan Keshav

PhD Project description: Peatlands account for just 2.84% of land area worldwide but provide a terrestrial carbon store around twice as large as global forest biomass. Within the UK, the East Anglian Fenlands offer one of the highest per area mitigation opportunities in the country. However, a current lack of low-cost, accurate and scalable methods for assessing the costs and benefits of regeneration projects is crippling potential investment. Advances in remote sensing technologies and machine learning algorithms offer a new and exciting solution to this problem. My  project will investigate how these methods can be applied to improve the trust, economic viability and ultimately uptake of peatland regeneration projects. The first objective is to gain an improved understanding of how peat depth varies across the Fens. Secondly, the factors affecting the economic cost of restoration will be explored. Finally, a regeneration prioritisation map of the Fens will be produced, highlighting areas where the greatest societal cost-benefit ratio can be obtained through regeneration.

Before joining the CDT, I graduated from Imperial College London, where I completed my undergraduate degree in Electrical and Electronic Engineering. Having completed multiple industrial placements within various electronics departments, including within a formula one team, I decided that the CDT was the best opportunity for me to make a difference to combating climate change, while providing me with the skills I need for a career in artificial intelligence. In my spare time, I like to keep active and am willing to give most sports a go - including hockey, football, snowboarding and most recently surfing. 


Thomas Hojlund Dodd


PhD project title: Bayesian optimisation led design of biopolymer-based thermoset composites with enhanced recyclability

Supervisors: Prof James Elliott (Dept. of Mat. Sci. & Met.), Prof Rachel Evans (Dept. of Mat. Sci. & Met.), Prof Alexei Lapkin (Dept. of Chem. Eng. & Biotech.)

Before arriving at Cambridge, I completed an MSc in the study of Energy Systems at the University of Oxford, where I wrote my thesis on the modelling of hybrid renewable energy systems and their integration with hydrogen electrolysers. This followed three years at the University of Manchester where I obtained my BSc in Environmental Science; specialising in the geochemical modelling of interactions between low-temperature fluids and high-strength rock. As part of Cambridge’s AI4ER CDT programme’s 2021 cohort, I finished my first-year MRes with a project on the optimisation of carbon mineralisation in basaltic rock using a sequence of reactive transport model simulation, gradient boosted decision tree emulation, and Bayesian optimisation. Moving from the Department of Earth Sciences to the Department of Materials Science and Metallurgy for my PhD, I currently study the potential for Bayesian optimisation to accelerate the development of sustainable properties in thermoset plastic materials. Current work is split between looking at improving recyclability and attempting to increase the quantities of sustainably derived feedstocks used.

Madeline Lisaius

PhD project title: Self-Supervised Learning Approaches in Earth Observation Leveraging Physical Knowledge about the Interaction of Electro-Magnetic Radiation with Vegetation Canopies as Applied to Agriculture

Supervisors: Professor Srinivasan Keshav, Dr Clement Atzberger (Mantle Labs)

Before joining the CDT, I was the Lead Data Scientist for topics of remote sensing, food and climate and Tom Ford Fellow at The Rockefeller Foundation. I graduated from Stanford University in 2019 with a B.S. with Honors in Earth Systems and an M.S. in Earth Systems with a focus on remote sensing for applications in macro land systems, especially agriculture. During my time at Stanford, I spent four years collaborating with indigenous women in the Ecuadorian Amazon as a National Geographic Young Explorer understanding working on projects of forest disturbance, women's empowerment through new land management regimes and the acute limitations and opportunities of Earth Observation for community-lead management. I am interested in advancing applications of remote sensing to topics of food, climate and gender as well as broadening critical discussion of Earth Observation and it's relationship to oppression and injustice.
In my PhD, I am working on adapting methods in self-supervised learning to leverage Earth Observation data for applications in agriculture.

Jonathan Roberts

PhD project title: Large-scale perception and visualisation of environmental data using 'few-shot' learning

Supervisors: Dr Samuel Albanie, Dr Kai Han (University of Hong Kong), Professor Emily Shuckburgh (CST)

PhD Project description: My PhD project focuses on using computer vision and natural language processing techniques to understand satellite imagery. More specifically, we are looking to leverage the broad knowledge of foundation vision-language models to robustly interpret remote sensing data with minimal training. I joined the AI4ER CDT as part of the 2021 cohort. I completed a Master of Physics degree at the University of Warwick, where I undertook research projects on the biophysics of microorganisms, global sustainable energy, and the orbital dynamics of geosynchronous satellites. After graduating, I worked as a Systems Engineer at Lockheed Martin for 3 years, conducting primarily modelling and simulation work across a diverse range of different projects and domains.

Before joining the AI4ER CDT as part of the 2021 cohort, I completed a Master of Physics degree at the University of Warwick, where I undertook research projects on the biophysics of microorganisms, global sustainable energy, and the orbital dynamics of geosynchronous satellites. After graduating, I worked as a Systems Engineer at Lockheed Martin for 3 years, conducting primarily modelling and simulation work across a diverse range of different projects and domains.

Sofija Stefanovic

PhD title: Monitoring impacts of exploratory mining with community citizen science and AI systems

Supervisors: Professor Alan Blackwell (Department of Computer Science and Technology), Professor Jennifer Gabrys (Department of Sociology)

PhD Project description: This research will endeavour to develop computer software for processing data related to environmental measurements such as water and soil pollution. The project will include research into novel algorithms, visualisations and analysis methods that apply artificial intelligence methods to environmental risk. The goal is to produce interactive software tools that enhance environmental monitoring. The methods will include design processes from the field of human-computer interaction, to ensure effective usability in the local contexts where environmental data is being collected. Supervised and unsupervised machine learning methods will be applied to remote sensing imagery for identification of ‘pollution fingerprints’, combined with in-situ sensor data to understand pollution patterns and automate parts of the monitoring process.

Before joining AI4ER, I spent a year doing a PhD at Oxford as part of the Marie Curie Innovative training network focused on applying machine learning to study aerosol-cloud interactions, which I left to focus on more human-centred research. I am interested in remote sensing and data science applications for environmental citizen science and organising in communities disproportionately impacted by climate change, pollution and environmental degradation. I completed my MSc in theoretical physics at Oxford in 2019, and BSc in liberal arts and sciences at Amsterdam University College where I majored in physics. Outside of the CDT, I work with a citizens' initiative I co-founded in Serbia focused on environmental monitoring, advocacy and education.

Sophie Turner

PhD title: Optimising Atmospheric Photolysis Simulations for Climate Models

Supervisors: Professor Alex Archibald and Dr Luke Abraham (Chemistry)

PhD Project description: My PhD project is Optimising Atmospheric Photolysis Simulations for Climate Models. Photolysis is one of the most computationally expensive components of the United Kingdom Chemistry and Aerosols model, which is part of the Met Office Unified Model and the Earth System Model. The aim of this project is to reduce the time taken for the photolysis simulation to run. This will be done by emulating the photolysis simulation, using a neural network. I’m doing this project because I enjoy computer science and I want to help improve climate models so that more informed decisions can be made in response to climate change.

Before joining the AI4ER programme, I studied computer science at University of Plymouth.