Getting the best out of probabilistic flood forecasting – on the art and science of aiding decision making during incidents

The Environment Agency (England & Wales)  recently commissioned research in the area of “Applying probabilistic flood forecasting in flood incident management” (Project: SC090032). Detailed reports can be found here.

Project Summary

What did the project deliver?

A set of reports have been produced which illustrate how probabilistic flood forecasts could be used to help decide on whether to take action when a flood is forecast. The information is mainly aimed at professionals and decision makers responsible for forecasting and responding to the risk of flooding.

Flood forecasts are uncertain. Unlike conventional, deterministic forecasts, probabilistic flood forecasts try to quantify and represent this uncertainty to enable longer forecasting and warning lead times and support risk and impact based decision making . They also help us to better understand the range of possible outcomes so we can better communicate the potential impacts on the ground. However, the additional information provided by probabilistic forecasts raises important questions, such as:

• “What action should be taken when some of forecasts predict flooding, while others do not?”
• “How can we best use probabilistic information to inform operational decisions which often require clear choices (e.g. whether to operate certain structures or not)”

This R&D project provided underpinning evidence, illustrative approaches and case studies to highlight how probabilistic forecasts could potentially be used to aid decision making in Flood Incident Management in future.

Our work confirmed that decision making during flood events is influenced by many factors ranging from ‘hard’ evidence (forecasts, observations and data) to important ‘softer’ factors such as local knowledge, recent flood history and forecast performance and current risk appetite. It highlighted that decision making during incidents is a dynamic process which can and should vary depending on the specific situation.

How can I use this Research?

To account for this, this research developed a flexible, yet simple framework to aid decision making with probabilistic forecasts which allows taking account of softer factors. It includes a number of possible methods (proof of concept stage), from the very simple to more complex, illustrated through practical case studies. The report and illustrative guide explain how these methods could be applied in principle to a variety of forecasting situations of different complexity and at varying lead times ahead of a potential flood.

The research makes an important contribution to our evidence base and provides a valuable resource to practitioners and researchers alike. It will help inform the future direction of probabilistic flood forecasting as part wider developments in flood incident management.

It is important to note that decision making with probabilistic information is an active research and development area and we expect novel approaches to become available over time. The proof of concept approaches described here should therefore be seen more as illustrations how probabilistic flood forecasts could be potentially used to support decision making and not as fixed and definitive operational procedures to be followed.

Detailed reports can be found here.