Professor Ian Vernon

Professor in the Department of Mathematical Sciences

What do you do?

I am a Bayesian statistician specialising in Uncertainty Quantification (UQ) for expensive computer models of complex physical systems.

My research interests include developing UQ methodology and applying it across multiple scientific disciplines including cosmology, epidemiology, systems biology, geology, nuclear physics and even to Bayesian statistics itself. This work dramatically improves our ability to understand such systems, our ability to predict them and our ability to perform robust decision making, whilst respecting all the many uncertainties that are present.

How are you involved in this area of science? 

I have worked on a series of projects that involved the collaboration of statisticians from Durham University’s Uncertainty Quantification group with leading scientific groups across a range of high profile areas. For example, I worked with Richard Bower and Carlos Frenk on their galaxy simulations known as Galform and EAGLE. We were able to calibrate their models by constructing emulators that mimicked their models but which were over 10 million times faster to evaluate.

This strategy has worked in multiple areas e.g. I also analysed models using emulators in systems biology, disease modelling and in nuclear physics. My main role involves thinking about and setting up the emulators and the larger uncertainty framework that the analysis requires. This involves assessing several key uncertainties that make the whole analysis trustworthy and robust.

What do you love about this topic?

There are many things I love about this topic, but first must be the ability to mimic an extremely expensive physical model (from cosmology, climate, epidemiology etc.) with a statistical construct known as an emulator, which is many orders of magnitude faster to evaluate. The opportunities that this then creates for scientists who have previously been greatly limited by the very slow runtime of their models is immense, and they are usually stunned as to its ramifications. The second thing I love is working with talented teams of scientists and statisticians: the interdisciplinarity of these projects is such a strength.

How does this work deliver real-world impact?

The impact of this work is readily apparent. Our analysis embeds expensive models within a robust uncertainty analysis leading to fast and efficient model calibration, prediction and decision support. This has been applied to many real-world examples e.g. the WHO commissioned a report on the worldwide rollout of TB vaccines, that utilised our work (also used for HIV, Covid and HPV). Our techniques were used in climate science and have influenced government climate strategy, and have been used in systems biology for plant root growth, DWP for pensions modelling and in the recent discovery of a new form of Oxygen.

Find out more:

• Discover how this work supported the discovery of a new form of Oxygen.