Avalanche approaching: James Cusick explains how new forecasting techniques could save lives on the ski slopes

On the Weissfluhjoch mountain above the fashionable ski resort of Davos in Switzerland, a small international research centre is seeking to develop forecasting techniques that will unlock the secrets of one of nature's most powerful events - the avalanche. A fortnight ago, five British doctors and their ski guide lost their lives in an avalanche in France. Before the winter is over, avalanches are likely to claim many more lives.

The Federal Institute for Snow and Avalanche Research is headed by a former nuclear physicist, Othmar Buser. The physics of snow, especially the mass of snow and ice that overcomes frictional resistance on a sloped surface, is now Dr Buser's business.

Although avalanches can occur during heavy storms when snow is still falling, most take place when the enormous power stored in accumulated snow is released. Weak spots occur in the packed snow near ground level and these act as a sort of lubricant but it is the upper layers of a snowpack, usually having less dense ice crystals, that begin sliding down mountains.

The traditional method of studying avalanche potential and evaluating danger, according to Dr Buser, focused on building up snow profiles and gathering weather data. Such profiles in most of Europe's main ski and climbing areas date back 50 years. But the emphasis then switched to computerising comparative statistics. The Davos institute is attempting to marry computer programming techniques based on human knowledge of avalanches with the more established comparison methods. The result, still some way off, is expected to represent a leap forward in forecasting techniques.

NXD, or 'nearest neighbours', is the name the institute gives one of its most commonly used techniques. The model incorporates information on: the comparative weight of new snow; wind direction and velocity, thus measuring snow drift; comparisons in temperatures through the snowpack; and air temperatures. Note will also be taken of the angle of a particular slope, the shape of the terrain, its orientation towards sun and wind and the altitude of the mountain slope.

Given that a winter in Switzerland lasts around 270 days, and the collected data spans 20 years, the number of 'comparison' days available to the institute is 5,000. The 10 days nearest to the 'measured' conditions are delivered from the data bank. If an avalanche occurred on three of those 10 days, then the model estimates there is a 90 per cent chance of an avalanche occurring. Essentially, 'nearest neighbours' is an evaluation of any slope's inherent stability, given the histories of similar slopes in similar conditions.

The Scottish Avalanche Information Service (SAIS) in Aviemore feeds mountain statistics from Scotland into the Swiss data bank. The SAIS also uses the 'nearest neighbours' model based on data it has gathered over the past five years. Valuable research on the effects of rain on snow, a phenomenon not too well-known in Switzerland, but familiar to Scottish skiers and climbers, is now sent to Davos.

Another computer program, Avaloc, using the knowledge of mountain experts to build up slope profiles, was also developed at Davos. Past data, crucial to NXD, is not required. Instead hundreds of 'expert' rules are fed into a multi-factorial program that also produces a stability index.

Using both programs a ski resort can produce an avalanche warning index of one to five (where five represents the highest risk). High-risk areas in a resort can be 'cleaned' of avalanche hazards with explosive charges causing controlled slides.

Dr Buser said: 'Although this sounds like a lot of data has to be gathered, the systems have been designed so that they can run on small portable computers.'

A French version of an advanced NXD system, known as 'Adiclhima' is operated in Tignes, scene of last week's avalanche. The French also have a complex program that models the evolution of a snowpack and takes in national weather and snow statistics. The system, known as Crocus, has to be run on a large mainframe computer.

Working alongside Dr Buser is Robert Bolognese. He aims to merge NXD and Avaloc into a new system that will utilise artificial intelligence, the expert model, and past statistics, into one forecasting package. Early results of the system, named NX-Loc, according to Dr Buser, are encouraging. But there is a strong element of pragmatism in the way he views science and its ability to deal with the avalanche. 'We have more new ideas, but we are never going to be able to calculate this as we would one of Newton's laws. It is not like a pencil falling down.'

What would he like to see happen? 'What we should be able to do is to go inside a dangerous slope to do the measurements. But you don't go there because it's dangerous, do you?'

(Photograph omitted)