Quantitative risk assessment and risk management processes are critically examined in the context of their applicability to the statistically infrequent and sometimes unforeseen events that trigger major disasters. While of value when applied at regional or larger scales by governments and insurance companies, these processes do not provide a rational basis for reducing the impacts of major disasters at the local (community) level because in any given locality disaster events occur too infrequently for their future occurrence in a realistic timeframe to be accurately predicted by statistics. Given that regional and national strategies for disaster reduction cannot be effective without effective local disaster reduction measures, this is a significant problem. Instead, we suggest that communities, local government officials, civil society organisations and scientists could usefully form teams to co-develop local hazard event and effects scenarios, around which the teams can then develop realistic long-term plans for building local resilience. These plans may also be of value in reducing the impacts of other disasters, and are likely to have the additional benefits of improving science development, relevance and uptake, and of enhancing communication between scientists and the public.