7 February 2014

The Seti River debris flow in Nepal – what was the role of the smaller landslide downstream?

Posted by Dave Petley

A reminder about the Seti River debris flow

In May 2012 a debris flow swept down the Seti river in Nepal, taking 72 lives and causing considerable damage.  Long term readers will remember that a day after the debris flow I suggested that there might be a landslide cause, and that a few days later, thanks mainly to help from Colin Stark, we identified that a large landslide was likely to have been the trigger.  Over the next few days, and thanks to help from numerous people, we pieced together the the sequence of events.  Although there were those that disagreed with our interpretation at the time, our initial analysis has subsequently proven to be correct.  This does make it somewhat galling to see a NASA article entitled “One Scientist’s Search for the Causes of the Deadly Seti River Flash Flood“.  One scientist? Not at all – this was a huge team effort.

Anyway, an interesting aspect of that article, by Jeff Kargel, is the suggestion that the debris flow was caused by a secondary landslide downstream that had blocked the valley, creating a lake.  In this theory, the rock avalanche that swept down from the flanks of Annapurna IV triggered the breach of the landslide dam, releasing the destructive debris flow.  The NASA article maintains that this was a significant factor:

However, the more we searched, the more it became evident that this was definitely not a GLOF, but was caused by a rockslide into the Seti River gorge, formation of an impoundment reservoir over a several week period due to damming of spring snow and ice melt, and then the final triggering event of the mighty rock and ice avalanche off Annapurna IV.

Odd wording given that we had already clearly shown that this was not a GLOF, but there you are.  Anyway, does this idea of a landslide dam being the source of the water and sediment stack up?  Well, the first thing to note is that this landslide was certainly present in the gorge.  The scar is visible in the Landsat 7 ETM+ imagery that NASA collected a few days before the landslide, and an image taken a few days before the Seti River debris flow also shows it clearly.  This is the image collected before the rock avalanche:

Seti River

I have highlighted the landslide in red on the image above – it is clearly visible.  For reference the cliff that collapsed to generate the initial rock avalanche is the area highlighted in yellow. I’m afraid that in my view the evidence supporting the idea of a landslide dam and lake runs out from this point.

Kargel’s idea is that, as we identified, the rock avalanche sent a proportion of debris down the very steep gullies into the main channel.  This debris then caused the failure of the landslide dam, generating the debris flow.  The above image was collected on 20 April 2012, about a fortnight before the rock avalanche.  Colin and I looked at this imagery at the time, and we maintained that there was no need for this smaller landslide to play a role.  I retain that view.

The first problem that I have with the Kargel model of a valley blocking landslide theory is the lack of evidence for a lake in the satellite imagery.  Lets zoom in to the section just upstream of the landslide itself – again this is the 20 April 2012 (i.e. before the rock avalanche) image:

Seti River


Of course we get into resolution problems (hence the blurred nature of the image), but I see no evidence at all for a water body upstream of the blockage.  Perhaps it is hidden in the shadows, but if so it could not be a very substantial body of water, surely?

The second problem I have with this is that for a breach event to release a catastrophic debris flow we should see a huge breach scar in the landslide debris – remember that this is a catastrophic collapse event. Well, Dhananjay Regmi has a presentation online that includes a helicopter image of the landslide:

Set River

Dananjay Regmi: http://issuu.com/johnharlinmedia/docs/regmi_investigationsetiriverdisaste?e=7165924/4998805


To me this does not look like the aftermath of a catastrophic breach event.  Maybe there has been post-breach modification of the channel?  But actually there is no evidence of lake deposits in the image either.

The third aspect that I find hard to understand is the downstream morphology of the channel in the post-debris flow NASA image.  This is the view of the channel downstream of the landslide a few days after the Seti River debris flow:

Seti River..

I have highlighted the landslide in red.  There is a key point to understand about landslide dam breach events.  This is that the flood wave should be at its largest at the dam crest.  As the wave passes down the channel from that point it attenuates – reduces in height.  So, the impact should be at its maximum immediately below the dam, and this should steadily diminish downstream.  Does that happen in the image above? I hope that you can see that the exact opposite occurs.  The impact on the channel clearly gets larger and wider downstream – compare just below the landslide with the yellow circle – downstream the flood has clearly become larger, and even more so by the time it reaches the orange circle, and it gets even bigger thereafter.  This is not a dam breach flood.  Actually, this is the signature of a debris flow that is accumulating volume en route down the valley.

Still don’t believe me? Well, OK, I’ve saved the best for last.  You see when Captain Maximov videoed the rock avalanche he also caught this smaller landslide in the film.  He captured an image of it on his first orbit, which is about the time the landslide started.  This is a still from the video:

Seti River..

The landslide is directly below the nose of the plane, upstream is to the right. There is no lake, even though this was taken as the rock avalanche initiated..

So what happened?  The increase in size of the debris flow scar downstream suggests that the debris flow was entraining its bed (over-running and incorporating river bed sediments) as it travelled downstream. This is a very well-established process in debris flows, and it is not exotic.Most importantly, it explains all of the observations outlined above.

So the downstream landslide did not play a major part in this tragedy, and this was not a dam break flood.