Uncertainty quantification to understand environmental models

Lindsay Lee (University of Leeds)

Alan Turing Frank Adams 1,

I will introduce an application of UQ in earth sciences and the challenges remaining.  Focussing on the climate effect of global aerosol as one of the persistent dominating uncertainties in the Intergovernmental Panel on Climate Change (IPCC) report I will summarise current community efforts to quantify and constrain uncertainty.  I will then present work carried out in Leeds using statistical methods including expert elicitation, experimental design, emulation, sensitivity analysis, cluster analysis and history matching of a global aerosol model where uncertainty quantification is an integral part of the scientific process.  I will show that while the goal has been to assess uncertainty in the modelling focus we have actually gained more knowledge of the model behaviour on a process-based level as well as a better understanding of how observations can be used to constrain the model.  We have also been able to propose reasons why the uncertainty in the climate effect of aerosol has remained persistent despite improved scientific understanding, increased observations and increasing model resolutions.  I will finish by talking about how we are moving forward with uncertainty quantification in the aerosol community and the challenges that remain. This talk will be application focussed and accessible to non-mathematicians.      

Import this event to your Outlook calendar
▲ Up to the top