Uncertainty modelling in multi-criteria analysis of water safety measures

• DOI: 10.1007/s10669-013-9442-9
• Publication Date: 2013-05-06T07:50:31Z
• Authors: Andreas Lindhe; Lars Rosén; Tommy Norberg; Jon Røstum; Thomas J. R. Pett...
• Keywords: 13. Climate action; 0208 environmental biotechnology; 02 engineering and technology; 6. Clean water; 12. Responsible consumption; 3. Good health

Water utilities must assess risks and make decisions on safety measures in order to obtain a safe and sustainable drinking water supply. The World Health Organization emphasises preparation of Water Safety Plans, in which risk ranking by means of risk matrices with discretised probability and consequence scales is commonly used. Risk ranking enables prioritisation of risks but there is currently no common and structured way of performing uncertainty analysis and using risk ranking for evaluating and comparing water safety measures. To enable a proper prioritisation of safety measures and an efficient use of available resources for risk reduction, two alternative models linking risk ranking and multi-criteria decision analysis (MCDA) are presented and evaluated. The two models specifically enable uncertainty modelling in MCDA and they differ in terms of how uncertainties in risk levels are considered. The need of formal handling of risk and uncertainty in MCDA is emphasised in the literature and the suggested models provide innovations that are not dependent on the application domain. In the case study application presented here, possible safety measures are evaluated based on the benefit of estimated risk reduction, the cost of implementation and the probability of not achieving an acceptable risk level. Additional criteria such as environmental impact and consumer trust may also be included when applying the models. The case study shows how safety measures can be ranked based on preference scores or cost-effectiveness and how measures not reducing the risk enough can be identified and disqualified. Furthermore, the probability of each safety measure being ranked highest can be calculated. The two models provide a stepwise procedure for prioritising safety measures and enable a formalised handling of uncertainties in input data and results.