A new modeling system created by NASA researchers is the first of its kind to allow scientists to predict the risk of landslide activity in any part of the world in “near real-time.”
As detailed in a news release from NASA, the space agency developed the prediction model at its Goddard Space Flight Center. The system uses rainfall, which was described as the leading cause of landslides around the world, as the main variable that could point to the risk of such events taking place. Should conditions below the surface in a certain area be considered unstable, massive rainfall could potentially serve as a catalyst of landslide activity, resulting in a combination of rocks, mud, and/or debris tumbling down hills and mountains.
“Landslides can cause widespread destruction and fatalities, but we really don’t have a complete sense of where and when landslides may be happening to inform disaster response and mitigation,” said Goddard landslide expert Dalia Kirschbaum, who co-authored a paper published in the journal Earth’s Future detailing the new model and how it works.
“This model helps pinpoint the time, location and severity of potential landslide hazards in near real-time all over the globe. Nothing has been done like this before.”
A report from UPI summarized the methodologies used by the Goddard Space Flight Center when it created the new model, stating that the system blends data gathered from satellite images and predictive analytics to come up with its near real-time predictions. In order to analyze historical landslide activity in a given area, the model makes use of machine learning techniques, while NASA and Japan Aerospace Exploration Agency (JAXA) satellites provide the required rainfall data, almost in real-time, allowing for the most updated and relevant risk data available.
In addition to the above data, the model also uses a global susceptibility map in the event that precipitation levels are “unusually high” in a given area. According to NASA’s news release, the map takes five variables into account, all of which are important in predicting the possibility of landslide activity — the presence of nearby roads, trees having been burned or removed, the proximity of a “major” tectonic fault, weak bedrock in the area, and steep hillsides. If, based on these five variables, the model points to a location with heavy rainfall as being vulnerable, it then produces a “nowcast” that warns of high or moderate chances of a landslide taking place. New notifications are sent out every 30 minutes thereafter.
In a statement, study co-author and Goddard landslide expert Thomas Stanley said that the new model has the potential to help scientists understand the risks of imminent landslides “in a matter of minutes.” Given that the model also proved to be effective in analyzing trends when the new data was compared to historical information, he added that it could be the first of its kind used to provide a retroactive snapshot of landslide activity over a certain period of time.