Climate extremes, floods and modelling approaches

Global warming is expected to increase the likelihood of extreme precipitation events and flooding. This seminar brings together two leading experts – Oxford University’s Dr Fredi Otto, and Professor Rob Lamb, Chief Scientist of JBA Consulting – to discuss modelling advances and the implications for flood risk management and climate adaptation.

The seminar will feature presentations from both speakers followed by a Q&A. Abstracts for each talk are found below:

Dr Fredi Otto: What is driving the rain? An event attribution approach to flooding
Some consequences of a warming world are predictable, such as an increased likelihood of more heat and high precipitation extremes. However, these so-called thermodynamic consequences only hold on average globally. Regionally, impacts can be influenced by a number of factors including dynamic effects of anthropogenically induced changes in atmospheric circulation, and local forcings.

To understand how the risks relating to extreme flood events are changing requires holistic perspectives, which combine an understanding of both thermodynamic and dynamic effects, and can translate precipitation estimates into flood forecasts.

Recent advances in the emerging science of event attribution provides the necessary basis on which to build the tools needed to assess such risks at the spatial scales that matter to people.

Professor Rob Lamb: Statistical models for floods of all shapes and sizes
Planning for the risk posed by flooding involves learning from what has happened in the past, but also being prepared for events that could be worse in the future. This means that a risk analysis should take account of plausible extreme events that encompass historical observations, and also of events that could be more extreme than anything previously seen.

Applications of extreme value theory have a long history in the analysis of river flows and are widely used in engineering practice and in flood risk planning to estimate the probabilities of extreme flood events.
Up until recently, statistical methods employed both in research and in practice have been suitable for analysing extreme events at specific locations, but not able to assess the joint probability of extreme conditions being experienced at multiple locations. Yet real flood events come in many shapes and sizes. A risk assessment for extreme flood events should therefore consider the joint probability of extreme conditions over many locations and multiple scales.

This talk will discusses a statistical modelling approach for multivariate extremes that has been applied to model the joint probability of extreme river flows over the entire network of around 1,000 river flow gauges in England and Wales. The model accounts for both the statistical distribution of extreme river flows at each gauge and the spatial and temporal dependence between gauges. This model has been used within insurance industry catastrophe risk models, new emergency planning scenarios for the National Risk Assessment, national scale assessment of the risk of bridge failures on the British rail network and to inform the government’s 2016 National Flood Resilience Review.