21 December 2018

Landslide probability may depend more on riverside steepness than on hillsides above

Posted by Lauren Lipuma

By Diego Arenas

On April 25, 2015, a 7.8-magnitude earthquake stuck the Gorkha region of Nepal near the capital city of Kathmandu. Approximately 9,000 people died and more than 22,000 suffered injuries. The quake also triggered more than 20,000 landslides in the surrounding area.   

A team of scientists at the University of Southern California is studying how the topography of the Melamchi Valley in Gorkha affected the incidence of landslides after the 2015 earthquake. Hazard models predict steeper land slopes are more likely to cause landslides, but when the researchers charted how the hillslope angle and frequency of landslides changed from south to north in the valley, they found the two did not align. While the hillslope angle steadily increased across the latitudes, the number of landslides remained constant before spiking in the northern portion of the range.

Before-and-after photographs of Nepal’s Langtang Valley showing the near-complete destruction of Langtang village due to a massive landslide caused by the 2015 Gorkha earthquake. Photos from 2012 (pre-quake) and 2015 (post-quake) by David Breashears/GlacierWorks.

“That was a little bit head-scratching,” said Joshua West, an assistant professor of Earth sciences at USC who presented the research last week at the 2018 AGU Fall Meeting in Washington, D.C.

West and his team found a stronger correlation when they juxtaposed the trend in landslide density to the steepness of the riverside. “Interestingly, it seems like how steep the river channels are is tracking with the landslides more so than how steep the hillslopes are,” he said.

Outside the valley, landslide density was also consistently more sensitive to channel steepness. “If you zoom out, this is one of the more visually striking relationships,” West said.  On a map of Gorkha featuring the channel steepness and landslides locations after the earthquake, landslides bunch around the branches with the steepest gradients. “It’s pretty striking that at the landslides really cluster around all these steeper channels to the north,” West said.

Whether the relationship between landslides and river channel steepness exists for other earthquake-prone zones remains to be determined, according to the researchers. They are planning a similar study for the Kaikoura earthquake, also of 7.8 magnitude, that stuck the South Island of New Zealand in 2016, but they expect more difficulty in obtaining data comparable to that of Nepal.

Shaking intensity during the earthquake and rock strength are varied relatively little throughout Melamchi Valley, which allowed the scientists to control for those factors and focus on the effects of terrain slope alone. “The nice thing here is we’ve been able to isolate the variables in a way that’s hard to in a lot of other past earthquakes,” West said. 

Although the research results don’t define the causal link between steeper channel slopes and increased frequency of landslides, West ventures possible explanations. He speculates that “if you’re cutting down more steeply, maybe you’re cutting underneath the rocks on the hillslope so that the rock mass above fractures differently.”

Alternatively, “by cutting down faster, you change the way water flows,” West said. Changing how water courses in the river changes the way it interacts with rocks and potentially weakens them.

Regardless of how exactly the channel steepness affects landslide frequency, West sees implications for his work. “If you were to use a linear dependence on slope angle, you wouldn’t really capture what we’re observing. It’s something that we should think about incorporating in hazard models.”

Marin Clark, a geologist at the University of Michigan who was involved in the new study, is doing just that. She is working on factoring the team’s data to develop models that can be implemented immediately after an earthquake.

“In the immediate aftermath of an actual earthquake, we have more specific information that comes within minutes and we’re able to take that specific information and rerun the hazard model for a scenario that just happened,” Clark said. “That scenario is helpful because it helps prioritize response efforts.”

-Diego Arenas is a student in MIT’s Graduate Program for Science Writing. Follow him on Twitter at @_D_Arenas.