23 May 2012
Using satellite imagery and aerial photographs to understand how a landslide caused the Seti River flood
By David Petley and Colin Stark
The Seti River debris flow of 5th May 2012, which is believed to have killed 72 people, was triggered by a landslide on the flanks of the Annapurna mountain chain. In our earlier posts, we highlighted that Colin, working with Goram Ekstrom, had detected this landslide using seismic instruments, and had been able to evaluate its characteristics. Furthermore, the landslide was both observed by Captain Maximov of Avia Club Nepal, and captured on a video camera mounted on the wing of his aircraft.
Over the last couple of weeks we have been working to try to understand this landslide. We have been helped by a wide range of people and as a result now have a reasonable understanding of what we think happened. This post presents our initial findings.
First, Captain Maximov and Avia Club Nepal have been exceptionally helpful in providing more information about the landslide. The following three images are used with their permission but are copyright of Avia Club Nepal. The first shows the landslide as it occurred. The slide is clearly generating a huge cloud of dust – this is an important observation in terms of our interpretation below (note that we have rotated and cropped this image to get the best possible view of the landslide):
These two images show the deposition area of the landslide. They were taken a few days after the event:
Whilst the obvious feature in these images is the large depositional zone on the surface of the lower angled slopes, if you look carefully you can see that on the steep slope behind there is a scar on the mountain side that extends to the ridgeline. We have zoomed in on this area in the image below, and have tried to change the contrast to show the feature more clearly:
Note in particular the section of the snow cornice that is missing in the top right side of the image. Once you have seen this the fresh scar is really very clear. So it is clear that the landslide is a huge rockslope failure – note that this cliff is about 2000 m high, on the flanks of Annapurna IV. The mass appears to have fragmented when it reached the lower angled slopes at the foot of the steep section, and to have runout as a rock and debris avalanche.
To better constrain this phase of the landslide, we have been working with Michon Scott of NSIDC, and Robert Simmon and Jesse Allen of NASA, who have been kindly collecting satellite imagery for us. They will be posting a summary of their work on the NASA website shortly (and I will update this post to provide a link), but their help has been invaluable. The best image was collected on 6th May by the Landsat ETM+ instrument. Unfortunately there are technical problems with Landsat that mean that the images have missing data, which generates the tiger-stripe effect that you can see below. However, the data quality is fantastic, such that the jey features of the landslide are quite clear. So here is the unannotated image, and below we have included one that highlights the key features of the landslide:
So, as shown above our interpretation is that the landslide consisted of a main detachment event from the steep 2000 metre high rock slope flanking Annapurna IV. This rockslope failure had a volume of about 22 million cubic metres. At the toe of the steep slope the mass fragmented and transitioned into a rock avalanche that flowed down the very steep slop. As the gradient started to decline deposition was initiated. The landslide generated a series of flows – based on previous events it is reasonable to think that there may have been a series of collapses following the main event that generated some of the lobes seen in the imagery. The fragmented mass generated a huge volume of dust, which was carried northwards by the wind to form the large aeolian deposit shown in the image above.
However, there is one remaining oddity that we have yet to resolve fully. The image above shows that the landslide did not reach the main channel – indeed even the most distal lobe is some considerable distance from main channel. So how did the landslide generate the huge debris flow that travelled down the main channel? At the moment we can only speculate, but the most likely explanation is that a small proportion of the debris entered and then flowed down one or more of the steep gullies that descend about 2000 m to the main channel. The most likely candidate is visible in the Landsat ETM+ image:
We have highlighted the most likely candidate gully, though close inspection of the image shows that there are other candidates, and it may well be that the flow travelled down several gullies at the same time. So, we hypothesise that a small proportion of the landslide travelled down the gully system, and entered the main channel at high velocity. This would then have entrained debris and fluid to create the destructive debris flow that claimed so many lives.
Finally, on 20th May NASA also imaged the Seti River from the landslide source (now covered in snow) down to Pokhara using the ALI instrument:
We have more work to do on this landslide, but as you will see from the above we now have a quite good understanding of it. We would welcome your comments and thoughts.
An interesting point to note is just how closely the analysis of the seismic data is to our interpretation of the seismic data is to the actual event as shown in the imagery. The power of this technique is becoming clear.