Assessing global change when data are sparse

Natural disasters and large‐scale industrial accidents are rare events having very low probability of occurrence. They may result in high consequences for society, the environment and the economy. Media reports often suggest that the occurrence and severity of such hazards is perceived to be changing. Change may be a consequence of global environmental changes such as global warming, tectonic and geological changes, or changing human activity. The assessment of global hazard change using standard statistical methods can be challenging due to sparse or lacking data leading to increased parameter uncertainties. To compensate for the sparseness of data, a hierarchical model approach is introduced where similar hazards (or groups of hazards) are combined within one model. Statistical inference must then 'borrow information' from similar hazards resulting in informed risk analysis and decision making. As an example in this paper, the hierarchical approach is applied to the question: "Have tsunamis becoming more or less frequent in the last century?"