De Kleermaeker et al. A decision support system for use of probability forecasts Proceedings of the 10 th International ISCRAM Conference – Baden-Baden, Germany, May 2013 T. Comes, F. Fiedrich, S. Fortier, J. Geldermann and T. Müller, eds. 290 A decision support system for effective use of probability forecasts Simone De Kleermaeker (i) Deltares, Delft, The Netherlands (ii) Ministry of Infrastructure and the Environment, Water Management Centre of the Netherlands, Storm Surge Forecasting Service, Lelystad, The Netherlands simone.dekleermaeker@deltares.nl Jan Verkade (i) Deltares, Delft, The Netherlands (ii) Ministry of Infrastructure and the Environment, Water Management Centre of the Netherlands, River Forecasting Service, Lelystad, The Netherlands (iii) Delft University of Technology, Delft, The Netherlands jan.verkade@deltares.nl ABSTRACT Often, water management decisions are based on hydrological forecasts, which are affected by inherent uncertainties. It is increasingly common for forecasters to make explicit estimates of these uncertainties. Associated benefits include the decision makers’ increased awareness of forecasting uncertainties and the potential for risk-based decision-making. Also, a more strict separation of responsibilities between forecasters and decision maker can be made. A recent study identified some issues related to the effective use of probability forecasts. These add a dimension to an already multi-dimensional problem, making it increasingly difficult for decision makers to extract relevant information from a forecast. Secondly, while probability forecasts provide a necessary ingredient for risk-based decision making, other ingredients may not be fully known, including estimates of flood damage and costs and effect of damage reducing measures. Here, we present suggestions for resolving these issues and the integration of those solutions in a prototype decision support system (DSS). A pathway for further development is outlined. Keywords Probabilistic forecast, predictive uncertainty, hydrology, decision-making, decision support system. INTRODUCTION Hydrologic forecasts are affected by inherent uncertainties, both epistemic and aleatory in nature. It is increasingly common for forecasters to make explicit estimates of aleatory uncertainties, i.e. to produce probabilistic forecasts (Cloke and Pappenberger, 2009). Associated benefits of probabilistic forecasts include the decision makers’ increased awareness of forecasting uncertainties and the potential for risk-based decision- making, as described further on in this paper. Also, a more strict separation of responsibilities between forecasters and decision maker can be made. These benefits can only be realised if the decision-response stages of a forecast-decision-response system are designed to take probabilistic forecasts as input, i.e. if the probabilistic forecasts are used effectively. PROBLEM DEFINITION In the scientific literature, some evidence is shown that using the probabilistic forecasts in risk-based decision making can reduce the users’ long-term flood risk, even though these forecasts are unlikely to have ‘perfect skill’ (Verkade and Werner, 2011; Zhu et al., 2002). However, simply having probabilistic forecasts available is not sufficient to realise the associated benefits. Additional effort is required in areas such as forecast visualisation and communication as well as making effective use of the probability forecast in decision-making processes. Forecasters and decision makers need to work together to develop strategies of how to resolve these issues (Nobert et al., 2010).