Uncertainty propagation through an atmospheric dispersion and radiological assessment chain of models

Public Health England (PHE) is participating in a number of work packages (WP) within an EU funded project titled CONFIDENCE (COping with uNcertainties For Improved modelling and DEcision making in Nuclear emergenCiEs). WP1 of the project aims to propose practical approaches to model improvement in the pre-release and release phase of a radiological accident, through uncertainty analysis with an ensemble approach. A comprehensive literature review was undertaken to identify and characterise the main types of uncertainty which are likely to impact on the modelling of atmospheric dispersion in the event of an accidental release to atmosphere. Based on this review, guidelines for ranking uncertainties in atmospheric dispersion and for detailing the range and distribution of atmospheric dispersion model input parameter uncertainties were developed. The uncertainties identified and characterised in the literature review were subsequently used to inform an uncertainty propagation exercise through atmospheric dispersion and radiological assessment models, for both historical events and hypothetical scenarios. Within this project (as would happen in a genuine emergency response) PHE worked in close collaboration with the UK Met Office (MO), applying atmospheric dispersion model output (including uncertainties) from the MO’s NAME model. However, contrary to PHE’s modelling approach in response to an accidental release, NAME model output was applied in PHE’s Probabilistic Accident Consequence Evaluation (PACE) software to evaluate the radiological consequences of the scenarios. The aims of this work include understanding which uncertainties are most significant in emergency response assessments, in part by considering if they are likely to impact on decisions (for example the implementation of countermeasures); identifying which uncertainties can and cannot realistically be considered in emergency response; and recommending an efficient and robust method of deriving uncertainties, notably by way of ensembles.