Catchment scale strategies are difficult to appraise and visualize because measures both engineered structures and features that work with natural processes to mitigate floods could be deployed in many different geographical configurations, and exposed to uncertain future weather events. Furthermore, predictions of the effects and benefits of some proposed features, especially within natural flood management are uncertain, and likely to remain so for some time.
A realistic evaluation of risk reduction benefits delivered by a catchment-wide strategies therefore needs to include uncertainties. More and more, we rely on chains of meteorological, hydrological, hydraulic and impacts models to inform decisions, yet models are sensitive to assumptions in parameters controlling process rates. These can be constrained better only if we incorporate data from many sources, but to do this effectively, also requires better ways of visualizing model responses in terms of multiple outputs under a wide range of scenarios.
Drawing on research funded by the European Commission 7th Framework and Defra/Environment Agency joint programme, we present a practical approach to catchment risk analytics, inspired by modern data science methods. This uses web-based tools to help analysts interpret the cost-effectiveness of catchment flood strategies using large ensemble simulations, in which uncertainty in both modelling and future extreme events is combined. The results are displayed as interactive graphs and maps. The approach, called UNCOVER allow for more informed decisions based on uncertainty, help to assess model performance across a whole catchment, appraise business information, and undertake trade-off analysis where a range of near-optimal solutions could be needed to reach robust decisions. It also provides tools to explore the benefits of different mitigation or monitoring strategies by data-mining results without the need for more simulations.Back to all speakers