Using Dempster–Shafer Theory to Represent Climate Change Uncertainties

When the prospect of climate change is admitted into the design and operation of long-lifetime water resource projects then the already high level of uncertainty is further increased. Currently, the longer-term climate is discussed in terms of a set of climate change scenarios. As a result of severe weakness in the knowledge and data pertaining to climate change, the relative probabilities of these scenarios are rarely indicated. This predominantly possibilistic rather than probabilistic view of climate change falls short of the needs of the established approaches to quantitative resource management. Weak information and its associated uncertainties are easily distorted when quantified and this can lead to decision analysis results that are misleading. A promising scheme for dealing with such information, based on Dempster–Shafer (D–S) theory, is described in this paper. Through the use of probabilities that are assigned to freely defined intervals of values, D–S theory offers greater flexibility than the Bayesian approach when quantifying weak information and more faithfully reflects its consequences in the results of decision analysis. At the same time the Bayesian approach and D–S approach share many fundamental ideas and produce identical results when the uncertainties are less extreme. This paper presents, along with some elementary examples, aspects of the D–S approach that contribute to its appeal when dealing with weak subjective and data-based information sources that have a bearing on climate change. The topics discussed include: the Basic Probability Assignment (BPA) and its visualization; combining BPAs from different sources; representing ignorance and near-ignorance; capturing subjective knowledge; inference; and decision analysis.