Research in risk modeling and evaluation in eWater Collaborative Research Centre has two emphases: (a) Integrating risk modeling and evaluation into the common decision making process of eWater so that eWater tools can be used to model and evaluate risks. (b) Providing risk measures that can be integrated into the eWater models. The study presented in this paper demonstrates that the process of risk modeling can be integrated into eWater's decision making framework, where performance data and risk data (variability and likelihood) are generated and used. It also demonstrates that the Value-at-Risk (VaR) method can be used to formulate risk-based objective functions, and that minimising risks can be treated in the same way as minimising values like costs and losses. It is expected that the next generation eWater tools will be extended to incorporate risk data, multiple-objectives assessment, and the Monte-Carlo simulation so that risk modeling and performance modeling can use results from each other. We conducted the case study in three stages. The first stage was to identify risks that were related to decision making in service delivery in the public domain (water resource assessment). We analysed the risk context in multiple domains. For this we followed the steps of a previously established, system-based risk management framework, to demonstrate the process of identifying risks and relating them to key system elements via the impact chain concept "control --> cause/factor --> system component --> risk" (--> means "influences"). Through these steps, we succeeded in relating risks with their corresponding factors and causes (decision variables) in a high level schematic diagram. The second stage of the case study was divided into two parts. The first part studied how the historical inflow sequence was used to perform water resource assessment. The second part adopted the historical inflow sequence (historical simulation) approach, but reformulated the problem into that of minimising a loss function Z(x, r(alpha)), by adjusting the decision variable x of general security (GS) water allocation, for the inflow sequence r(alpha) corresponding to an exceedance probability alpha. We introduced two measures Value-at-Risk (VaR(alpha)(x)) and Conditional Value-at-Risk (CVaR(alpha)(x)), borrowed from the financial sector, to measure risks. In the water domain, VaR(alpha)(x) is the alpha-quantile of the loss function Z(x, r(alpha)) for a water system, whereas CVaR(alpha)(x) is the conditional expected loss for Z >= VaR(alpha)(x). We noted that the approach offers at least two benefits: (a) The measures of VaR and CVaR associate water allocation risk with water inflows, supporting a better understanding and explanation of risks and actions. (b) Minimising risk is treated the same as minimising cost or loss; thus opening up the opportunity of applying stochastic optimisation methods in risk assessment. In the third stage we set up scenarios for a range of exceedance levels alpha = 0.99, 0.95, 0.90, 0.80, and 0.60. The optimised solution x was contingent on the inflows into the river dam. If the inflow was very low (i.e., alpha very close to 1), the GS allocation x could be very small. Conversely, if alpha was not so close to 1, e. g. 0.90, 0.80, or 0.60, then the GS allocation x would grow accordingly. While we appreciated the potential merit of measuring VaR and CVaR in economic and other values, we did not have time to pursue this in the study. Instead, we focused on demonstrating the feasibility of integrating risk management into decision making. In the future, we will expand the idea of setting up risk values as objective functions and applying stochastic optimisation methods to reduce risks.
