A step forward to predicting rare events

Update this image alt text please

 This week Nature Physics (03 May 2009) reveals how a scientist from the London Centre for Nanotechnology at UCL (University College London) has found a simple way of predicting the chance that a larger than normal event will happen. This will help us to understand the changes of river levels, and have application in the engineering of sustainable energy sources such as fusion reactors.
Many properties in physical and social science are loosely described as "critical" - from floods and earthquakes to the turbulence that affects aircrafts and poses a problem in engineering. Critical properties are often thought to have power law distributions (see picture), but this is only true for some properties: others typically show a very distinctive, but non-power law distribution. The first example of this was noted by Professor Steven Bramwell (UCL) with his colleagues Peter Holdsworth and Jean Francois Pinton (ENS Lyon) in connection with fluid turbulence. However, such  phenomena are so complex that it is usually very difficult to predict which sort of distribution will occur.
Now Professor Bramwell has enunciated a simple and general way to predict what type of distribution should be expected in a given practical situation such as floods and turbulence. If it is known how a parameter changes with the size of the system, then the rule may be used to predict the distribution of that parameter. The parameter could be a river level or the amount of fuel escaping from a fusion reactor.
Professor Bramwell says “This will help us understand rare events and will help us in making predictions for complex phenomena where a simple model is not possible.”
In a world of climate change and high technology, this new development will be a milestone in statistical research, allowing more efficiently forecasting of events that influence us all.
Notes for editors
Contact details:
For more information, please contact Dr Thierry Bontoux at the London Centre for Nanotechnology on tel: +44 (0)20 7679 0056, e-mail: t.bontoux@ucl.ac.uk
The following images can be obtained by calling the London Centre for Nanotechnology on +44 (0)20 7679 0056 or by emailing t.bontoux@ucl.ac.uk
A power law distribution (blue) versus the typical non power law shape (red).
The blue curve might represent how the chance of an earthquake diminishes with increasing earthquake magnitude, while the tail of the red curve might represent the chance of a large event in turbulence. 
About the London Centre for Nanotechnology
The London Centre for Nanotechnology, is a UK-based, multidisciplinary research centre forming the bridge between the physical and biomedical sciences. It was conceived from the outset with a management structure allowing for a clear focus on scientific excellence, exploitation and commercialisation. It brings together two world leaders in nanotechnology, namely University College London and Imperial College London, in a unique operating model that accesses the combined skills of multiple departments, including medicine, chemistry, physics, electrical and electronic engineering, biochemical engineering, materials and earth sciences, and two leading technology transfer offices.