Climate change is leading to increasing hydrological extremes and quicker shifts between wet and dry extremes in many regions. These extremes and rapid shifts put pressure on reservoir operations, decreasing the reliability of water supply, flood control and other reservoir benefits. Decision-makers across all levels, from reservoir operators to flood plain residents, turn to heuristics to simplify decisions when faced with complexity and uncertainty, resulting in cognitive biases or systematic errors in decision-making. While cognitive biases are not new, climate change is exacerbating their impact for two reasons: 1) heuristics, just as infrastructure, are based on experience with historic conditions; 2) fragilities created by these cognitive biases can go undetected until extreme events occur. If not acknowledged and managed, these cognitive biases can lead to catastrophic failures of reservoirs and other infrastructure. To minimize risk of such catastrophic failure, we propose a multi-level approach to flood and drought management, one that strikes a balance between centralized and decentralized approaches. Such an approach is better able to cope with uncertain and changing conditions because it creates overlaps and diversity, which can respond to a wide range of conditions and builds checks and balances that mitigate cognitive biases latent in various decision-making units.