Rosie Archer

My background is in Mathematics, specialising in Statistics. I am interested in applying statistics to real-world problems.

I am currently studying for my PhD within the PALGLAC project, working to create an optimal numerical model of the last Eurasian Ice Sheet using statistical techniques.

My other research interests include Bayesian statistics, machine learning and statistical distributions of landform morphology.

Emulating the Last Eurasian Ice Sheet

My project is looking at using statistical techniques, such as emulation, to improve approaches to modelling the last Eurasian Ice Sheet. Ice sheet models are computationally expensive, and produce large uncertainties. An emulator is a statistical representation of model simulations, which can be used to infer model behaviour for parameter combinations that have not been run through a model. This means that an emulator can help quantify uncertainties as well as reducing the computational power required. In my project, I will model the EIS using the Parallel Ice Sheet Model (PISM). Output from the model will be used to train a Gaussian Process emulator. The emulator will use model-data comparison techniques, to find how different parameter combinations alter the fit to palaeo data.

Current techniques of model-data comparison techniques are primitive in their statistical approach and rarely used. From the emulator, a Bayesian history matching approach will be used to further refine parameter values to use within the main ice sheet model. By improving the model-data comparison framework, my aim is to better utilise the record from palaeo-ice sheets to find optimal parameter combinations, which in turn may help improve ice sheet model formulations.

Supervisors: Dr Jeremy Ely, Dr Tim Heaton, Prof Chris Clark.