Science News

Predicting Landslides

Sun, 8th Mar 2009

Listen Now    Download as mp3 from the show Your Questions and Swallowing Swords

Predicting earthquakes and avalanches is notoriously difficult, scientists have been attempting to do so for hundreds of years with very little success, and a group from imperial college london may have worked out why.

Both earthquakes and avalanches are types of critical phenomina, the classic example is slowly pouring sand onto a sand pile. The top of the sandpile slowly gets more and more unstable until something gives and you get a landslide. The problem is predicting how big the landslide is going to be. It might be tiny or the small landslide, may trigger a larger one which may trigger an even larger one, or it might not.

Henrik Jensen has been looking at a simplified version of this. He has been creating a pile of ball bearings by adding one at a time to the top of the pile. Every time he added a ball he took a photo. Occasionally there were landslides of different sizes which he tried to predict in various different ways.

He found that it was virtually impossible to predict the size of the landslide the traditional way, by looking at the size of previous landslide. But he did have more luck when he looked at the state of the pile before each ball was added.

He found that the more disordered the pile was before the next ball was added, the larger the landslip was going to be. and he could predict the size of the landslip with a 64% accuracy, and he thinks that he can get a lot better.

This sounds like a pretty abstract finding, but it does show that if you are wanting to predict avalanches or earthquakes you shouldn't be looking at previous avalanches or earthquakes, but look at the state of the hillside or tectonic plates. And possibly more importantly it means that the problem is soluble if we approach it from the right direction.

References

Multimedia

Subscribe Free

Related Content

Not working please enable javascript
EPSRC
Powered by UKfast
STFC
Genetics Society
ipDTL