US states turn to lidar and machine learning to map landslide risk

The US Geological Survey has awarded $1 million to 16 state, local and tribal governments, including Pennsylvania, Nevada, North Dakota and Alaska, for landslide inventories, hazard maps and risk reduction, StateScoop reported.

There is no official count of landslides, but a University of Washington database launched in July shows about 6.5 million Americans live in areas at risk.

Lasers over the landscape

"Lidar was probably the biggest, most important tool that has revolutionized our ability to accurately and precisely delineate landslides," said Stephen Slaughter of USGS. Laser measurements reveal terrain changes hard to see in aerial photos. Deep landslides leave clear signatures; shallow ones after heavy rain "go faster than someone can run" and are the main threat to life.

Pennsylvania is building its first landslide database. A two-year grant will map landslides and produce a susceptibility map for Washington County, near Pittsburgh. In the past, staff examined topographic maps and aerial photos by hand. "We haven't had the tools or the staff," said Stephanie Evans of the state's geological survey. "It just wasn't physically possible when you only had like one or two people."

The state will use 2020 lidar data while a statewide survey updates coverage, so future comparisons can show whether known slides have moved. It is also exploring machine learning to flag likely landslide features automatically, an approach similar to California's Wildfire Commons, which pools hazard data from many agencies.

Small grants for data infrastructure can have large effects. A good susceptibility map shapes building permits, road maintenance and evacuation plans for years, and machine learning lets a team of two cover a whole county.