The construction of the Grand Ethiopian Renaissance Dam (GERD), a massive hydroelectric project on the Blue Nile, is part of the broader Nile River basin, which spans across 11 countries in northeastern Africa, including Ethiopia, Sudan, and Egypt. The GERD is designed to be the largest dam in Africa, with the primary purpose of generating electricity to power Ethiopia’s development and provide energy to neighbouring countries. However, its construction has ignited significant debate, particularly due to concerns from downstream countries, Sudan and Egypt, which rely heavily on the Nile's waters for agriculture, drinking water, and energy production. The dam's operation and management have far-reaching implications for the entire Nile basin region. Adding to this complexity is the uncertainty of future climate conditions, which could drastically change water availability in the region. To navigate these challenges, it is essential to develop robust reservoir management strategies—policies that are resilient and adaptable to various future scenarios. This thesis explores a broader and more detailed approach to evaluating the robustness of water management policies, focusing on the case of the Nile River. Instead of relying on a single metric, we use multiple robustness metrics; Percentile-based Skewness, Mean-Variance, Undesirable Deviations, and Minimax Regret to see how they might lead to different conclusions about the best strategies.

To evaluate the resilience of these policies against future uncertainties, a series of steps are involved. First, policies are generated using optimization techniques designed to identify the best strategies that could foster cooperation among the countries, while also addressing their individual objectives. These strategies are then tested for their effectiveness under various future scenarios, which is done by applying robustness metrics. A robustness metric is a quantitative measure used to assess the resilience and stability of a system or process in the face of uncertainties, disturbances, or perturbations. It provides a way to quantify how well a system can maintain its performance or functionality under varying conditions. These metrics can range from assessing absolute performance or regret prioritizing risk-aversion, maximizing performance, or minimizing variance, depending on the specific uncertainties and the decision-maker's risk tolerance. However, many previous studies done on robustness of reservoir control rely on the 90th Percentile Regret metric which looks at how much worse a given strategy performs compared to the best possible outcome. While useful and practical, this approach doesn’t fully capture the range of different ways to test against future scenarios.
This research reveals that the choice of robustness metric can greatly affect the evaluation of different policies. Using four different metrics, we were able to conclude different “most robust” policies. Using minimax regret metric, Compromise Policie(s) are the most robust. However, using Undesirable Deviations metric, Best Egypt Irrigation Policy is the most robust. Using Percentile-Based Skewness yields Compromise: Percentile and Best Sudan Irrigation Policy as the most robust, and using Mean-Variance metric, only Best Sudan Irrigation policy emerged as the most robust. This variation occurs because different robustness metric prioritizes different aspects of policy performance under uncertainty. While the regret metric focuses on minimizing the worst-case scenario outcomes, other metrics like Percentile-Based Skewness emphasize the consistency of outcomes across a range of scenarios.

These findings emphasize the importance of using a diverse set of robustness metrics in policy evaluation. A diverse set of metrics allows for a more comprehensive assessment of policies by capturing different aspects of performance and risk under uncertainty. Depending on which metric is used, different strategies may be recommended, leading to potentially different outcomes for the countries involved. While strategies may seem mutually exclusive, a more nuanced approach can involve balancing the trade-offs highlighted by the various metrics, leading to a more informed and robust decision-making process.