Determining the most informative scenarios of environmental impact from potential major accidents

In many situations where an agency is aware of potential disasters, it is necessary to analyse in some depth, possible disaster scenarios. The reason for analysing a scenario is similar to analysing in depth any example—it forces a precision and is likely to bring to mind aspects not realized when thinking in general terms. The example scenarios are also likely to form the basis for data collection, to «put some numbers» on the extent of potential damage and the probability of that damage. Furthermore, a scenario, like any example, serves as a basis of understanding and communication of the projected risks. Unfortunately the effort dedicated to developing and analysing a number of scenarios can be great, and such expenditure may not be justified. What one seeks is to identify one, or at most a few, potential accidents whose analysis gives maximum information about the total accident risk in many or all of its aspects, without demanding an excessive budget to develop and analyse a large number of scenarios. The paper considers probability, damage and similarity as appropriate bases for selecting those accidents on which can be built scenarios that are not only important in their own right, but will also give information about the results of other possible accidents. An integer linear program is formulated, with the objective of maximizing the similarity between the selected candidate accident(s) and all identified potential accidents. This is found to be computationally simple to solve with modern software. The approach is demonstrated by applying it to define a «most informative» scenario of a major liquid spill in an inland waterway.