Accounting for Model Error in Earth System Digital Twins

This PhD project, based at the University of Reading's Department of Mathematics and Statistics, focuses on advancing Earth system digital twins—cutting-edge computational models that integrate machine learning with real-world environmental observations. These digital twins are designed to predict the impact of policy choices on the environment and society, providing rapid, data-driven insights for decision-makers. The research sits at the intersection of applied mathematics, machine learning, and climate science, aiming to develop next-generation digital twins that rigorously account for uncertainties and systematic errors in complex physical models. Central to the project is the use of machine-learned emulators, which enable fast simulations of physical systems under varying conditions. These emulators are continually updated with observational data from satellites and ground-based instruments using data assimilation techniques, ensuring that model predictions remain closely aligned with reality. The project will address the challenge of model error—systematic discrepancies between physical models and real-world processes—which can propagate through machine learning emulators and affect the reliability of digital twin outputs. The student will develop new mathematical and computational methods to quantify and mitigate the effects of model error during both the machine learning and data assimilation stages. This involves theoretical work using numerical linear algebra, as well as computational testing on idealised systems before applying the techniques to more realistic climate models. The research is highly interdisciplinary, drawing on expertise from the Data Assimilation Research Centre (DARC) at Reading, the Data Learning Group at Imperial College London, and the UK National Centre for Earth Observation (NCEO), all of which are actively developing digital twins for climate applications. Training opportunities include seminars and courses organised by these research groups, as well as access to NCEO's annual training events and conferences for early career researchers. The project is fully funded for UK students, offering a UKRI stipend and home-level PhD tuition fees through the EPSRC Centre for Doctoral Training in the Mathematics for our Future Climate. Applicants should have a strong mathematics background and an interest in machine learning, climate science, or computational modelling. Applications are accepted year-round. For further details, visit the project page or contact the Department of Mathematics and Statistics at the University of Reading.