Robust Yellow River Delta Flood Management under Uncertainty

Abstract

A number of principles for evaluating water resources decisions under deep long-run uncertainty have been proposed in the literature. In this paper, we evaluate the usefulness of three widely recommended principles in the context of delta water and sedimentation management: scenario-based uncertainty definition, robustness rather than optimality as a performance measure, and modeling of adaptability, which is the flexibility to change system design or operations as conditions change in the future. This evaluation takes place in the context of an important real-world problem: flood control in the Yellow River Delta. The results give insight both on the physical function of the river system and on the effect of various approaches to modeling risk attitudes and adaptation on the long-term performance of the system. We find that the optimal decisions found under different scenarios differ significantly, while those resulting from using minimal expected cost and minmax regret metrics are similar. The results also show that adaptive multi-stage optimization has a lower expected cost than a static approach in which decisions over the entire time horizon are specified; more surprisingly, recognizing the ability to adapt means that larger, rather than smaller, first-stage investments become optimal. When faced with deep uncertainty in water resources planning, this case study demonstrates that considering scenarios, robustness, and adaptability can significantly improve decisions.