Dealing with Epistemic Uncertainties in Estimating Vulnerability and Risk
Analyses of vulnerability and risk very often make an intrinsic assumption that sources of uncertainty are aleatory in nature or can be treated as if they were aleatory. This is generally not the case for either data or models. The epistemic nature of many sources of uncertainty that arise in risk assessments mean that the effective information content associated with the misfit between observational data and model predictions might be much less than the aleatory assumption would suggest. This is why over-confident decisions might be made and why there is often an element of surprise in actual outcomes relative to risk predictions. Methodologies for dealing with epistemic uncertainties necessarily rely on more subjective expert judgments than formal statistical analysis.
It is hoped that this mini-symposium will be an opportunity to discuss some of the issues raised by epistemic uncertainties, what their impact is on the information content for model calibration, and methodologies for allowing for them into decision making particularly in mitigating the risk of natural hazards.