Dr Andrew Meijers, Prof G Vallis
No more applications being accepted
Competition Funded PhD Project (European/UK Students Only)
About the Project
The Southern Ocean surrounding Antarctica is the principal region where the deep ocean, cryosphere and atmosphere freely exchange properties with one another. This is the major pathway for heat, carbon and nutrients into the ocean interior and has a disproportionately large impact on global climate. However, such exchanges of active tracers within a complex dynamical system creates difficult-to-predict coupled feedbacks that may profoundly influence climate. Presently climate models, including the UK’s national model, do not produce coherent future projections for the Southern Ocean, representing a major source of uncertainty for global predictions (Meijers 2014).
Recently emerged machine learning (ML) techniques have a great, and largely untapped, potential to improve our analysis of climate models and our understanding of the wider climate system. This project seeks to apply these exciting new ‘climate informatics’ (Monteleoni et al. 2012) to the problem of understanding how polar change will impact global climate.
This project will utilise emerging AI tools to examine a suite of state-of-the-art climate models and characterise the key dynamical relationships setting the state of the polar climate. It will link these causal relationships and our understanding of the observed ocean with possible future states of the Southern Ocean and use the relationships uncovered to reduce our uncertainty in global climate projections.
The student will apply ML algorithms such as dimensionality reduction, clustering and graphical networks (Ebert-Uphoff & Deng 2012) to ensembles of coupled climate models. Unsupervised learning techniques offer considerable potential to reveal relationships within models. Challenges include adapting existing algorithms to deal with the very high-dimension data and its non-stationary, non-isotropic and highly-correlated nature. The student will identify the key parameters in each model and contrast relationships across models to understand the role of ocean-atmosphere-ice coupling in setting future model states. They will constrain projections using the knowledge uncovered combined with known present-day properties and dynamics to reduce uncertainty in future global climate projections.
Funding Notes
This project would suit a numerate student with a background in mathematics, physics, data science, machine learning or an equivalent quantitative field, and a desire to apply new techniques to climatic questions. This work is data driven so experience coding and dealing with large datasets will be an advantage.
References
Meijers, A.J.S. “The Southern Ocean in the Coupled Model Intercomparison Project phase 5” Phil. Tran. R. Soc. A, (2014): 20130296
Monteleoni, C., et al. “Climate Informatics” in Computational Intelligent Data Analysis for Sustainable Development (2013): 81-126
Ebert-Uphoff, I. and Deng, Y. “Causal discovery for climate research using graphical models” Journal of Climate (25) (2012) 5648-5665