9 May 2012
Using seismic data to analyse the Seti River landslide in Nepal
Posted by Dave Petley
Colin Stark and Goran Ekstrom have been working for a while on the analysis of seismic data to detect very large landslides. Whilst still in its infancy, this technique shows great promise for very large landslides, and it is increasingly clear that some landslide parameters can be extracted from the seismic data if the force history inversion is undertaken correctly.
On 5th May 2012 the global seismic network detected what appeared to be a landslide event to the north of the town of Pokhara in Nepal. Soon after a flood wave travelled down the Seti River, killing over 50 people. It is likely that the wave was triggered by the collapse of a large landslide mass that had impounded the river.
Colin and Goran have now undertaken Landslide Force History (LFH) inversion for the event that they detected on 5th May, and to quote Colin (with his permission) “we are now 100% sure there was indeed a seismogenic slope failure in the Pokhara area”.
Their best estimate for the landslide parameters are as follows:
- Date/time: 5th May 2012 at 03:24:56 GMT = (i.e. 09:09:56 local time)
- Runout duration: ~103 seconds
- Max force: ~1.2*10^11 N assuming a mass of 1.4*10^11 kg (see note#1 below)
- Runout distance: ~1040 m (see note#1 below)
- Height drop: ~350 m (see note#2 below)
- Max accel: ~0.85 m/s^2 (depends on mass assumption)
- Max speed: ~24 m/s (ibid)
Colin notes that:
Note#1: There is an inherent ambiguity in the LFH inversion since we obtain force, which is mass*acceleration, which integrates up to mass*distance. If we specify mass, we predict runout distance (and height drop); if we specify runout distance or height drop, we predict mass. In this case I would bracket (generously) thusly:
- n/c
- n/c
- n/c
- Mass range: 5*10^10 kg — 2*10^11 kg
- Runout distance ~ 2910 m — 740 m
- Height drop ~970 m — 250 m (see note#2 below)
- Max accel ~ 2.4–0.6 m/s^2 (depends on mass assumption)
- Max speed ~ 67–17 m/s (ibid)
From experience I would guess somewhere in the middle is most likely, i.e., my guess above at a mass of around 1-2*10^11kg. Once we have satellite imagery, the bracketing will improve markedly.
Note#2: The inferred height drop is dependent on the mass assumption (or measurement), but it is apparently also dependent on slope failure mechanism – on occasion, we appear to underestimate (in the LFH inversion) the vertical force involved. This may simply be because our inversion scheme does not at present apply extra geomorphic constraints beyond the requirement of net stationarity. We’re working on this.
One of the most intriguing aspects of this technique is that it generates a trajectory of the centre of mass of the landslide. This is a planform diagram in which the length of the arrow indicates the relative speed:
One very useful result: the planform trajectory of the landslide center of mass. Here’s a pic (the arrow size gives relative speed):
“This landslide (centre of mass) moved almost directly westwards with a modest bend in its trajectory”
So, assuming that the detected event did provide the debris that triggered the collapse, where did it happen? Well, the analysis of the data suggests that it was in this area, althoigh it is not well constrained:
From there we can turn to the second amazing part of this story. Kunda Dixit is a well-known journalist in Nepal. A day or so ago he made the following comment on this blog:
The latest death toll is 17 dead 47 missing. The cause was neither glacial lake outburst nor ice avalanche but the collapse of a rockface on the eastern flank of Machapuchre.
You may find these interesting:
http://nepalitimes.com.np/blogs/kundadixit/2012/05/05/avalanche-flood-on-seti/
and
http://nepalitimes.com.np/blogs/kundadixit/2012/05/05/eye-in-the-sky/
Machapuchere is shown in the image above. However, the landslide moved from east to west, which suggests that it was perhaps a failure on the rockface downstream from Annapurna II. Note though this is just speculation at this time, and it could of course be that the landslide event that was detected and the event that caused the flood are unrelated (although the coincidence would be surprising).
Finally, it is worth noting that Kunda Dixit’s post points out that the alarm about this flood was raised by a light aircraft pilot, Captain Alexander Maximov, who saw it from his aircraft. There are even images of the front of the flow taken from the aircraft:
[…] the Seti River, close to Pokhara in Nepal, which killed 72 people. We are trying to tie together the amazing data from the seismic network with observations and timings. The primary source of information has been the remarkable videos […]
[My comments in square brackets – Dave]
This is false interpretation of the cause of SETI River Disasters. The landslide damming story is completely wrong because the debris has travelled a distance of 17 km from the source (i. e., the avalanche zone) in about 40 minutes [agreed].
We conducted an aerial survey recently, but found no traces of any pounding of water or debris. So, the sole cause is debris-mixed avalanche, which seems to have muddified/slurrified immediately after failure, probably because of the frictional movement of the debris-mixed ice blocks.
[OK, but check out the NASA satellite imagery, and remember that we have clear data on the nature of the lanslide from the LFH analysis of the seismic dataset]
Another possibility is that the avalanche also forced-broke the debris-mixed ice mountains (which we can still see in the source zone of SETI River, which also turned into debris slurry within a few minutes. We interviewed the Russian pilot, who told us that he saw the huge avalanche cloud around 9:00 AM, 5th May. The photo taken by the picnicking college boys indicates the time of their photo to be 9:38 AM, which means the debris travelled up to Kharapani (Tatopani) in just about 40 minutes. How come we can think of landslide damming or an LDOF? We are simply disseminating misleading information without doing necessary groundwork.
[But we are not invoking a barrier lake]
Now, we are also talking of another rubbish by raising the issue of seismically induced landslide or avalanche. There are chances that an imperceptible seismic event took place and induced the avalanche, but this particular event may rather be a geological process, which may be typical to the Himalayas and snow-covered mountains. What may happen is because of the continuous snow cover as well as variation in temperature, the slope material undergoes deterioration in a long period of time, say about 50-100 years. As the snow cover may go as high as a few tens of meters, the pressure on the deteriorated slope material rises causing possible creep of the slope material together with the glacier movement. After a certain time, particularly after physical force imbalance, some part of the slope or the snow cover fails and causes a heavy slide of the weakened part of the slope or snow cover. This may happen when the iced mass of snow cover is close to melting point, i. e., less than 0 degree Celsius. We found that on 19 May, the melting line was somewhere around 4,000m of altitude, which indicates that the snow cover around this altitude is close to melting point. On top of that, when some foreign materials are mixed in the ice, during sliding or movement, the ice may melt faster due to frictional heat. This might have happened in this particular case. However, this speculation also requires some experimental verification, which we plan to study further in the days ahead.
[Who is suggesting that this was semically triggered? Not us! It was not
seismically-triggered – it generated
a seismic signal]