PROJECT OUTLINE

PROJECT AIMS : vision and impact

HOME PAGEProject NewsPROJECT OUTLINE About Uncertainty in Models Background Project Aims Outputs of Project Facts Themes People & Structure CONTACTS DISSEMINATION Papers & Presentations Events DownloadsRESOURCES ToolkitReading ListRelated Projects Glossary MEMBERS ONLY Project Team Management Board Advisory Panel WIKI

Modelling is fundamental to research across all areas of science, industry and decision-making, yet uncertainty in model predictions is poorly understood at best; more often it is just ignored. The potential impact for both modellers and model users of having a technology that is accessible, and can routinely quantify and analyse uncertainty, even in complex computer-intensive models, is enormous. It will become possible for the first time to validate computer intensive models in a principled statistical framework, to compare models and to gauge their adequacy for specific purposes. For  instance, it will for the first time be possible to say whether a complex model is genuinely better than a simpler one that represents reality less accurately but is less sensitive to misspecification of its inputs. We will be able to examine the role of ensembles of models in gauging model structure uncertainty. It will also become possible for the first time to quantify the value of model calibration and data assimilation in improving the accuracy of model predictions. We will even be able to target research or observational studies to address unacceptable levels of uncertainty, and to predict the ways in which such research will change uncertainty.

For example, Rolls-Royce have a model for an aero-engine which can assess the risk of failure dependent on the geometry and tuning of the rotor blade assembly, temperature, possible damage to blades, etc. The run-time of the model is substantial; success in this project would allow them to quantify more efficiently the risks of failure, allowing for uncertainty about all those inputs, thereby shortening the design cycle. Furthermore, it would be possible to allow for uncertainties in the model itself and to assess the impact of the various sources of uncertainty. This will potentially enable them to gauge the benefits of tighter engineering tolerances, or of procedures to reduce the risk of blade damage.

Tools to quantify, analyse and understand uncertainties in models comprise a basic technology, but to realise their potential they must be accessible routinely to modellers and end users across the whole domain of scientific modelling. There are big technical challenges in making the step change from the current state of the technology, but the basic science is done, the potential for a dramatic impact is demonstrated and a consortium of researchers, who already work together and with modellers and model users across a wide range of disciplines, is in place.