Following shortly after his recent contributions to seismic monitoring and nuclear non-prolifereation, Professor Erik Sudderth of Brown University's Department of Computer Science and his collaborators have developed graph algorithms that use remote sensing data to predict where landslides are most likely to occur. Their work was published in the Journal of Geophysical Research: Earth Surface and was later chosen as a highlight article on EOS.
Landslide prevention presents a massive computational task: hillsides are typically modeled as grids, often composed of millions of cells or blocks, each of which has properties such as soil depth, slope, and elevation. The endless possible permutations make determining how unstable cells are arranged enormously demanding, so Sudderth and his collaborators developed an algorithm that analyzes the properties of hillsides and identifies clusters of unstable blocks, which in turn allows analysis of a larger area. They tested their work on a virtual hillside and on data from an Oregon landslide, and in both cases, scientists were able to accurately predict the area and approximate side of the landslides.
For more information, click the link that follows to contact Brown CS Communications Outreach Specialist Jesse C. Polhemus.