Using Dempster-Shafer theory to model uncertainty in climate change and environmental impact assessments

We present a methodology based on Dempster-Shafer theory to represent, combine and propagate statistical and epistemic uncertainties. This approach is first applied to estimate, via a semi-empirical model, the future sea level rise induced by global warming at the end of the century. Projections are affected by statistical uncertainties originating from model parameter estimation and epistemic uncertainties due to lack of knowledge of model inputs. We then study the overtopping response of a typical defense structure due to (1) uncertain elevation of the mean water level and (2) uncertain level of storm surges and waves. Statistical evidence is described by likelihood-based belief functions while imprecise evidence is modeled by subjective possibility distributions. Uncertain inputs are propagated by Monte Carlo simulation and interval analysis and the output belief function can be summarized by upper and lower cumulative distribution functions.