Reporter 424, 5 October 1998
A new computer-based flood warning system could protect homes and businesses from rising waters despite having the thought power of a brain-damaged worm.
Flooding in parts of Southern England earlier this year caused millions of pounds of damage, mainly due to the speed and surprise with which it occurred.
The artificial neural network developed in the Department of Geography could help buy vital time for those living in vulnerable areas, by accurately warning when a river is likely to burst its banks.
Based on the processes thought to occur within a biological brain, the neural network uses river level data to predict flooding and can progressively improve its performance to give more accurate forecasts.
The systems network of 24 neurons compares poorly with the billions found in a human brain, or even a worms 1000, but early indications suggest it can offer cheaper and more reliable data.
The projects leader, Professor Stan Openshaw believes such an approach is needed to tackle the effects of global climatic change.
Neural nets have the advantages of flexibility and speed of application. At a time when weather patterns are more extreme, this will become increasingly important, he said. The research team says that trials have shown the system to be consistently more accurate than current forecasting techniques and have just launched a web site to prove it.
The web interface allows river engineers to try it with their own data, explained Dr Simon Corne, and compare our predictions with their own.
The current system cannot be applied to cases of severe flooding, such as that from the giant rivers found in China and Bangladesh; but the team are working on a portable version that will be compatible with smaller scale rivers anywhere in the world.
[Main news stories | In the news | News in brief | Events | Notice board]
HTML by Jeremy M. Harmer