Improving the accuracy of breach modelling: why are we not progressing faster?

Flood risk assessment and management often requires the prediction of potential breaching of a flood defence embankment or dam in order to either assess potential impacts or provide information to assist emergency planning, evacuation, repair strategies and improve alternative future design strategies. There are many different aspects of the overall breaching process, which are more, or less, relevant to the wide range of potential end users of such information. Consequently, the prediction of breach growth is an area where research has been undertaken for many decades in an attempt to provide more reliable models and predictions. However, despite many initiatives providing observations and recommendations as to processes observed and how research might progress, more detailed literature searches will often uncover conclusions and observations noted a decade or two or three earlier that are similar to those being made today. In particular, observations relating to material type, state (such as water content and compaction) and properties are relevant here. This prompts the obvious question as to why our ability to predict breach initiation and growth has not progressed further over this period. Why are so many studies identifying similar issues and, in effect, 'reinventing the wheel'? With a programme of research into breach initiation and growth under the EC FLOODsite Project and continued pressure to improve tools and techniques following events such as those seen at New Orleans in August 2005, this paper considers this question of apparent slow progress and offers some suggestions as to why this may have occurred and what direction might prove more effective in the future.