University of Exeter: Maths & Statistics
Supervisor: James Salter j.m.salter @ exeter.ac.uk
Understanding how much carbon is absorbed by land ecosystems is essential for predicting future climate change and estimating the remaining carbon budget for limiting global warming. Scientists use computer models to simulate how forests, soils, and vegetation exchange carbon with the atmosphere. These models play an important role in climate research but contain many uncertain parameters that influence their predictions.
This project sits at the interface of climate science, statistics and machine learning (ML), and will contribute to ongoing research aimed at improving estimates of the global carbon cycle and carbon budgets. The project will use statistical and ML techniques to investigate how uncertainty in model parameters in the land surface model (JULES) affects predictions of carbon uptake.
The project will focus on using stats/ML tools to analyse a pre-prepared set of JULES simulations, to explore how different parameter choices affect the model’s representation of the real-world. Within this, there are many interesting questions, and the exact objective of the project is flexible, allowing the student to focus on a particular scientific question (for example, site-scale carbon fluxes or future climate scenarios) or on methodological aspects of model emulation and calibration. Through this work, the student will gain experience applying statistical and ML methods to real-world environmental modelling problems.
The student will be embedded within an active research group working on the global carbon cycle and climate modelling. They will have the opportunity to attend group meetings and interact with researchers working on climate science, environmental modelling, and statistics. The project contributes to ongoing research programmes at the University of Exeter and the Met Office, and the student will gain insight into how environmental models are developed and used in climate assessments. There will also be opportunities to engage with wider academic networks (e.g. statistics and environmental modelling groups) and to present their work in an informal research setting at the end of the placement. This project would be particularly suitable for students interested in pursuing postgraduate research combining statistics or machine learning with environmental or climate science.
