Propagation of Uncertainty through a Hazard Chain
A problem brought to the 85th European Study Group with Industry by the Defence Science & Technology Laboratory.
Problem Description
Toxic substances released into the environment pose both an immediate and delayed risk to human health. When this release is in the form of a gas or vapour it is necessary to predict where the substance will disperse and deposit in the environment as this will allow a first responder to undertake appropriate mitigation strategies. To this end many organisations have worked to produce models which predict parts of this process. However, in order to produce an estimate of casualties that may result from exposure to the substance these disparate models must be tied together. We are looking to establish a methodology for propagating uncertainty through this modelling chain in order to undertake a casualty estimate.
Study Group Report
The Study Group have looked at the particular example of casualty estimation from airbourne dispersion and suggested two different potential solutions. If the structure of the models is sufficiently simple, and the number of degrees of freedom relatively small, a semi-analytical approach based on Bayes' theorem can be used. In the more general case, intelligent sampling methods can be used to gradually build a picture of likely outcomes.