Landslide rescuers to get help from rapid analysis of seismic data

When a dangerous landslide hits, rescuers scrambling to assist need to know exactly where to look – and now a new technique for analysing seismic data can do just that, by tracing the source of the disaster to within a few kilometres in mere seconds.

Existing methods can only narrow down locations to within tens of kilometres, says Stefania Ursica at the Helmholtz Centre for Geosciences in Germany. In remote areas, this can lead to delays if rescuers are sent to the wrong place. “That time loss can be crucial,” Ursica told a press conference at a meeting of the European Geosciences Union in Vienna last week.

Many countries have a network of seismic monitors to record activity related to earthquakes and volcanoes. This data can also be used to detect events such as landslides – the risk of which is increasing due to climate change – but the data from these kinds of events is much messier and harder to analyse than that from earthquakes, says Urica.

There are two novel aspects to her team’s approach, she says. The first is to analyse five different aspects of the seismic waveform to pick out from the noise exactly when an event occurred.

This information is then fed to a dozen or so mathematical agents that search for the location of the event, such as the initial rockfall leading to a landslide. They do this by estimating what waveform would have been produced if the event had happened in a particular spot and comparing it to the recorded waveform. If it doesn’t match, they try a different spot.

Each agent “moves” in a pattern inspired by the behaviour of animals, from the spiralling of a falcon to the long migrations of elephants, until they have collectively homed in on the most likely location of the event. The entire process takes only around 10 seconds and is much more accurate than previous approaches. “We have basically an order of magnitude [of] improvement,” Ursica says.

In addition to helping rescue efforts, it will help researchers locate events in remote areas where satellite data is unclear or unavailable, she says: “We can pinpoint events that we otherwise couldn’t see.”

The team plans to publish the details soon and make the code available for others’ use.