Abstract Global economic losses related to natural hazards are large and increasing, peaking at US$380 billion in 2011 driven by earthquakes in Japan and New Zealand and flooding in Thailand. Catastrophe models are stochastic event-set based computer models, first created 25 years ago, that are now vital to risk assessment within the insurance and reinsurance industry. They estimate likely losses from extreme events, whether natural or man-made. Most catastrophe models limit the level of user interaction, stereotyped as ‘black boxes’. In this paper we investigate how model fusion techniques could be used to develop ‘plug and play’ catastrophe models and discuss the impact of open access modelling on the insurance industry and other stakeholders (e.g. local government).

Abstract Society faces a future of unprecedented, extensive and rapid environmental change. The impacts of climate change and greater societal vulnerability will require far-reaching adaptations of behaviour and activity. To plan these, decision-makers require tools that will help them understand the extent and impact of natural hazards. These should take into account deterministic and probabilistic analyses of occurrence, impact, spatial distribution, background conditions and triggers that affect different hazards. Geographic information system (GIS)-based, 2D, models are easily understood by different users and are well suited to situations where data is plentiful, the historical record is relatively complete or where problems are simple. However, they have limitations. Advances in computer processing capacity have brought marked changes in how data can be manipulated and presented; this has allowed and driven an increasing desire to provide more detailed information about the spatial extent, temporal occurrence, triggers and impacts of geohazards. Future geohazard models based on 3D distributions of causative factors, including primary (e.g. precipitation) and secondary (e.g. groundwater) 4D processes, that determine the timing, scale and geographical distribution of events will rapidly evolve. They will increasingly integrate data from other disciplines, such as societal vulnerability, to develop risk models. Currently, geohazard models are constrained by inadequate data, a poor understanding of the interaction of processes, and cultural barriers such as inertia or intellectual property rights. The development of improved models, whether for planning purposes or management of crises, provides challenges, both in system development and in the communication of complexity and uncertainty with decision-makers.