The linear, ‘rationalist’ policy model is the principal means by which governments justify and evaluate policy decisions, despite its practical difficulties and the widespread criticism it has received when accounting for the complexity, uncertainty and divergence of opinions and values associated with contemporary policy problems. Our research, part of a NCCARF-funded project to integrate climate change adaptation and disaster risk management (DRM), demonstrates how rationalist ‘predict-then-act’ approaches promote unrealistic public expectations of DRM and a reactive approach to natural hazards overall. Examining institutional responses to three recent natural disasters across Australia, our research also reveals how rationalist policy making masks normative decisions behind technical ‘evidence’; over-relies on technical expertise, engineering and the reduction of exposure to natural events; while neglecting the types of social capital required when engineering provisions inevitably fail, or fail to provide the level of protection expected of them. We propose an alternative approach, in line with the pressing need for climate change adaptation and the practical difficulties of reducing uncertainties. By re-casting the existing Prevent, Prepare, Respond, Recover model of DRM in terms of a normative, incremental policy cycle, we argue that DRM can become more adaptive to future climates so that communities will be progressively better prepared for each new climate extreme. This approach focuses on managing uncertainties rather than reducing them and building resilience not simply through the reduction of hazard exposure, but through the reduction of community vulnerability, explicit consideration of normative policy priorities and increased community engagement in climate risk debates