Below is a list of DPhil (PhD) project areas within Climate Physics. If you are interested in any of the research areas listed, please contact the relevant supervisor directly. They will be happy to discuss potential project opportunities and answer any questions you may have.
Please note that some projects may be filled as applications are reviewed. We therefore particularly encourage candidates who are considering applying after the January deadline to contact prospective supervisors in advance to discuss available project options.
- Atmospheric composition, observations and modelling: Paul Palmer
- Cryosphere and polar processes: Andrew Wells
- Earth and planetary climate dynamics: Raymond Pierrehumbert
- Geophysical fluid dynamics: Andrew Wells, David Marshall, Raymond Pierrehumbert, Tim Woollings
- Machine learning for weather and climate science: Antje Weisheimer, Hannah Christensen, Milan Klöwer, Paul Palmer, Philip Stier
- Natural and anthropogenic drivers of climate change: Anu Dudhia, Don Grainger, Paul Palmer, Philip Stier, Raymond Pierrehumbert
- Ocean physics: Andrew Wells, David Marshall
- Physics of clouds and aerosols and their effects on climate: Don Grainger, Philip Stier
- Predictability of weather and climate: Antje Weisheimer, Hannah Christensen, Milan Klöwer, Myles Allen, Tim Woollings
- Remote sensing and earth observations: Anu Dudhia, Don Grainger, Neil Bowles, Paul Palmer, Philip Stier
- Response to natural and anthropogenic drivers of climate change: Antje Weisheimer, Myles Allen
- Tropical climate variability: Antje Weisheimer, Hannah Christensen
- Variability of atmospheres and oceans: Antje Weisheimer, David Marshall, Hannah Christensen, Myles Allen, Tim Woollings
- Weather and climate modelling: Antje Weisheimer, Hannah Christensen, Milan Klöwer
Project examples
The following projects are listed to give candidates with Overseas fee status an indication of the research opportunities available. Candidates with Home fee status are encouraged to apply through ILESLA or the CDT in Intelligent Earth.
The following projects are available for the 2027/28 academic year:
- Determining Atmospheric Composition from Infrared Satellite Measurements -
Supervisor: Anu Dudhia
Viewed from space, the Earth's infrared spectrum (15-5microns) shows major absorption features associated with water vapour and carbon dioxide (the major greenhouse gases) and ozone. Since 2007, satellite instruments have been making routine global observations with high-spectral resolution which reveal the spectral signatures of a number of additional molecules with concentrations of parts per million or less. These include other long-lived greenhouse gases (CH4,N2O), molecules associated with air-quality (CO,NH3), volcanic emissions (SO2), vegetative activity (OCS, Methanol) and biomass burning (HCN,Ethane,Formic Acid), and many more.
The DPhil project would be to develop and improve algorithms used in the retrievals of these species, with a particular emphasis on exploiting the higher sensitivity measurements expected from the new IASI-NG instrument (launched in 2025).
- Differentiable atmospheric modelling: Learning from data between Earth’s and exoplanetary climates -
Supervisors: Milan Klöwer and Thaddeus Komacek
Atmospheric general circulation models are the backbone of climate models, being used to understand and predict climate change on Earth. Founded in physical laws, general circulation models can be generalised to exoplanetary atmospheres. While many atmospheric processes on Earth are well understood and accurately simulated as evident through the success of weather forecasting, some processes such as cloud formation and precipitation are less certain. Societally-relevant surface climate and extreme events like heat waves, however, strongly depend on those. At the same time, observations can constrain such uncertainties, improving climate predictions on Earth. On exoplanets this will reveal insights about the parameter regimes in which planetary habitability is possible. How to make atmospheric models automatically learn from observational data?
This project will build on top of SpeedyWeather, an atmospheric general circulation model written in the Julia programming language. SpeedyWeather is easy to use and extend, inspired by modern software engineering, covering functionality from data visualisation to high-performance computing. The candidate will continue to develop its differentiability through the automatic differentiation framework Enzyme. Differentiating through SpeedyWeather, we can optimise the unknown or less certain parameters determining the planetary surface climate and resulting habitability in the same way as neural networks are trained towards observations. And we can add neural networks to SpeedyWeather forming a hybrid physics and data-driven model.
For exoplanets, observations of gas giants down to hot terrestrial planets are currently used to constrain their atmospheric composition and thermal structure. Critically, upcoming space missions such as the Large Interferometer for Exoplanets and Habitable Worlds Observatory will have the ability to measure the thermal emission and reflected light of Earth-like exoplanets, enabling a direct test of our Earth-based understanding of planetary climate. This SpeedyWeather/Enzyme framework will enable novel rapid inverse characterization of exoplanetary atmospheres with a 3D model that can be applied to exoplanet observations.
Part 2 of this project will scale up the same method to the Earth’s atmosphere, increasing data and complexity. For earth, the current surface climate is observable, but a model may still mispredict frequencies and intensities of extreme events like heat waves under global warming. But with automatic differentiation we can train the model to correct for the missing physics of heat waves towards a more reliable generalisation into future climates, assessing the Earth’s habitability under global warming.
This DPhil project will bridge several fields: The applicant is expected to have a strong background on the spectrum between physics, mathematics, and computer science, with enthusiasm for computer science and machine learning. Prior experience with Julia is not required, but experience with another programming language like Python, Matlab, C(++), or Fortran is preferred.
- Projects on Atmospheric Composition -
Supervisor: Paul Palmer (joining the University of Oxford in 2027)
Atmospheric composition plays a central role in the Earth system through its influence on climate, air quality, and ecosystem health. Greenhouse gases such as methane and carbon dioxide drive climate change, while pollutants including ozone and particulate matter affect human health and agriculture. Despite major advances in observations and modelling, substantial uncertainties remain in emissions, atmospheric processes, and their response to natural and anthropogenic change.
Aims: Projects in this subject area will be developed jointly with the student and may focus on greenhouse gas emissions and sinks, air quality, atmospheric chemistry, satellite observations, inverse modelling, data assimilation, machine learning, or climate interactions.
Methods to be used: Depending on the specific project and student preferences, these questions can be tackled using atmospheric chemistry and transport models, inverse modelling and data assimilation systems, satellite and in situ observations, statistical methods, machine learning techniques, and high-performance computing, or any combination thereof.
Skills required: A solid quantitative background in Physics, Mathematics, Data Science, or a related discipline. Experience in programming (Python, FORTRAN), numerical modelling, data analysis, or machine learning would be advantageous but is not essential.
- How do large-scale climate anomalies influence methane emissions from tropical African wetlands? -
Supervisor: Paul Palmer (joining the University of Oxford in 2027)
Atmospheric methane is the second most important anthropogenic greenhouse gas and is increasing at rates that remain difficult to explain. Tropical wetlands are thought to play a major role in driving this variability, with emissions strongly influenced by rainfall, temperature, and hydrology. Large-scale climate anomalies, including El Niño events, provide a natural experiment to investigate how changes in climate affect wetland methane emissions. However, estimates of methane emissions from tropical Africa remain highly uncertain due to sparse observations and limited understanding of underlying processes.
Aims: This project will examine how climate variability influences methane emissions from tropical African wetlands using satellite observations, atmospheric inverse modelling, land-surface modelling, and machine learning techniques. An emphasis will be placed on understanding the influence of ENSO-driven changes in rainfall and hydrology, and on improving the representation of wetland methane emissions in Earth system models.
Methods to be used: Depending on the specific direction of the project and student preferences, the research will combine satellite observations (e.g. GOSAT, TROPOMI, CO2M), land-surface and atmospheric transport models (e.g. JULES and GEOS-Chem), inverse modelling, data assimilation, and machine learning approaches. The project will benefit from links to the NERC-funded CurFEW programme, which is developing new observational and process-level understanding of tropical wetland methane emissions.
Skills required: A solid quantitative background in Physics, Mathematics, Data Science, or a related discipline. Experience in programming (Python, FORTRAN), numerical modelling, data analysis, or machine learning would be advantageous but is not essential.
Further reading:
https://doi.org/10.5194/acp-18-18149-2018 McNorton et al. (2018) https://doi.org/10.5194/acp-19-14721-2019 Lunt et al. (2019) https://doi.org/10.1038/s41467-021-27786-4 Helfter et al. (2022) https://www.nature.com/articles/s41467-022-28989-z Feng et al. (2022) https://doi.org/10.5194/acp-23-4863-2023 Feng et al. (2023) https://www.nature.com/articles/s43017-023-00397-x Palmer et al. (2023) https://doi.org/10.5194/essd-17-1873-2025 Saunois et al. (2025)
- Revealing carbon cycle dynamics and their response to climate using satellite observations and AI -
Supervisor: Paul Palmer (joining the University of Oxford in 2027)
Atmospheric carbon dioxide and methane are increasing at unprecedented rates, reflecting changes in the global carbon cycle that remain only partly understood. Variability in terrestrial carbon uptake and release is thought to play a major role, particularly in tropical and high-latitude ecosystems, but the mechanisms driving these changes remain uncertain. Recent advances in satellite observations and AI provide an opportunity to identify the climatic drivers of ecosystem behaviour and improve understanding of how the carbon cycle responds to environmental change.
Aims: This project will investigate how climate variability influences terrestrial carbon cycling using satellite observations and emerging AI approaches. Emphasis will be placed on understanding ecosystem responses to large-scale climate anomalies, such as El Niño events, and on developing interpretable relationships between climate, ecosystem function, and atmospheric carbon dioxide.
Methods to be used: Depending on the specific direction of the project and student interests, the research will combine satellite observations of atmospheric composition, solar-induced fluorescence, vegetation dynamics, phenology, hydrology, and fire with machine-learning techniques including physics-informed neural networks, neural operators, and equation discovery. The resulting process relationships and predictive models will be evaluated using ecological measurements and compared with state-of-the-art land-surface and Earth system models.
Skills required: A solid quantitative background in Physics, Mathematics, Data Science, or a related discipline. Experience in programming (Python, FORTRAN), numerical modelling, data analysis, or machine learning would be advantageous but is not essential.
- Clouds in a Changing Climate: Earth Observations, AI and Global Kilometre-Scale Modelling -
Supervisor: Philip Stier
Clouds play a fundamental role in the climate system by regulating the Earth’s energy balance and hydrological cycle, but remain a major source of uncertainty in climate predictions. In particular, we need to understand how clouds respond to greenhouse-gas-induced warming through cloud feedbacks and to changing air pollution through aerosol-cloud interactions. New Earth observations, global kilometre-scale climate models and rapid advances in artificial intelligence now provide exciting opportunities to study clouds and their evolution in unprecedented detail.
Potential DPhil projects include:
1. Geospatial foundation model for clouds
Develop an observational AI foundation model that learns relationships between environmental conditions, clouds, aerosols and radiation from multi-modal Earth observations. The project could explore generative models for 2D/3D cloud reconstruction and evolution, physically structured representations of cloud regimes, or inference of poorly observed cloud and aerosol properties, with particular emphasis on physical interpretability and generalisation to changing climate conditions.
2. Calibration of global kilometre-scale models and aerosol-convection interactions
Develop innovative approaches to calibrate global kilometre-scale climate models using observations of cloud structure, lifecycle and radiative effects, combining perturbed-parameter ensembles with statistical and machine-learning emulators. Calibrated models will then be used to investigate aerosol effects on convection, precipitation and convective anvils using the ICON-HAM-lite modelling framework.
3. Lagrangian analysis of aerosol effects on convection using cloud tracking
Use the open-source tobac cloud-tracking framework developed in our group to follow individual convective systems through their lifecycle and determine how aerosol perturbations affect their development, precipitation, anvils and radiative effects. The project could combine satellite observations with regional and global kilometre-scale ICON-HAM-lite simulations.
4. AI nowcasting of extreme precipitation and haboob dust storms
Develop AI approaches for short-term prediction of deep convection, extreme precipitation and associated haboob dust storms, particularly in regions without extensive weather-radar networks. Building on our ConvCast concept, the project will combine sequences of geostationary satellite observations with atmospheric-state information to generate probabilistic nowcasts, with opportunities to explore prediction from convective precursors, uncertainty quantification, transfer to data-sparse regions and high-resolution simulations of storm evolution.
Aims: The precise scientific questions will be developed jointly with the student within one of these themes, spanning atmospheric physics, climate modelling, Earth observation and artificial intelligence.
Methods to be used: Depending on the project and student interests, methods may include global and regional high-resolution atmospheric modelling, satellite remote sensing, aircraft and ground-based observations, cloud tracking, statistical model calibration, computer vision, and modern machine-learning and generative-AI techniques.
Skills required: A strong quantitative background in Physics, Atmospheric Physics/Meteorology, Mathematics, Computer Science, Data Science or a related discipline. Prior atmospheric-science experience is advantageous but not essential for candidates with strong quantitative or computational skills. Programming experience is desirable.