30 April 2014

The Seti River landslide disaster in Nepal: two years on

The Seti River landslide

Almost two years ago, on 5th May 2012, a catastrophic debris flow swept down the Seti River in Nepal, killing over 70 people.  Working with various colleagues and friends through this blog, we quickly reconstructed the events that caused the debris flow; the key aspect being that the initiation was a large rock avalanche on the flank of the huge Annapurna IV mountain.

Since the landslide the reconstruction of events that we compiled has proven to be largely correct.  It is clear that the starting point of the sequence of events was a massive rock slope failure; that this turned into a rock avalanche that swept across the high altitude plateau; that the debris avalanche entered the gorge system in the Upper reaches of the Seti River; and that this turned into a debris flow that swept down the river, causing the massive destruction and loss of life.

One event or two?

There is one aspect of the Seti River event that I do not believe we got right at the time.  The great controversy about this event has always been the transition from rock avalanche to debris flow, and in particular source of the water.  The confusion has been the NASA satellite image, which appears to suggest that most of the landslide was deposited on the plateau, with almost none reaching the gully system.  This is the annotated Landsat 7 image that Colin Stark and I posted on this site two years ago:

14_04 Seti 1


A different, and I think correct, interpretation was proposed in a paper by Jorg Hanisch and colleagues in the Journal of the Nepal Geological Society (Hanisch et al. 2013).  They point out that the evidence on the ground suggests that the deposit shown in the satellite image above is actually the remains of a smaller secondary landslide.  The main event flowed across the plateau and entered the main channel, leaving almost nothing behind.  In fact, the Google Earth imagery that is now available of the site of the rock avalanche strongly supports this interpretation.  This is a perspective image of the site of the landslide from before the collapse, collected on 24th December 2011:

14_04 Seti 2


And this is the image after the landslide, collected on 12th July 2012:

14_04 Seti 3

The modification to the land surface caused by the landslide is very clear,  The landslide has removed almost all of the smaller topographic features, sch as crevasses and gullys.  In doing so it would have entrained large amounts of ice and snow, which of course would have melted to provide large volumes of water.  What is also clear from this image, as Hanisch et al. (2013) suggest, is that a large proportion of the landslide entered the main gully system, and thus formed the catastrophic debris flow that caused such devastation downstream.  Thus, there is no need to invoke a more exotic mechanism for generating the mass that formed the Seti River debris flow, such as a downstream valley-blocking landslide or drainage of cave systems.


Hanisch, J., Koirala, A. and Bhandary, N.P 2013. The Pokhara May 5th flood disaster: A last warning sign sent by nature? Journal of Nepal Geological Society, 46, 1-10.

Comments/Trackbacks (0)>>

28 April 2014

Upper Madi Hydroelectric Project – another fatal landslide at a major power scheme in the Himalayas

Upper Madi HEP

The Upper Madi (sometimes called the Super Madi) hydroelectric project in Kaski, Nepal is a 44 MW “run of the river” power scheme currently under construction, designed to generate electricity to help plug the shortage in supply in Nepal.  It is being built in a steep river valley to the northeast of Pokhara.  In essence the project is simple – a dam across the river upstream of a steep section of the valley diverts water from the river into a low gradient tunnel system through the mountain.  About 5 km downstream, where the difference in head between the tunnel and the riverbed is high the water descends down a steep section of tunnel and through a set of turbines before re-entering the main channel.  Thus, the key aspects of the construction are the upstream dam and associated infrastructure; the tunnel system; and the turbines and associated infrastructure; plus all the haul roads, cable routes, etc.

The Upper Madi landslide

On Friday a landslide occurred at the portal to one of the tunnels being constructed as part of the Upper Madi project.  The aftermath of the landslide is shown in the image below, from Republica:


The landslide trapped 15 workers within the tunnel.  Of these, 12 were rescued over the weekend, but the other three lost their lives.

It is a little difficult to work out exactly what happened here, other than a rockslope failure that impacted on and apparently collapsed the first few tens of metres of the tunnel.  However, this image, from FNN, appears to offer some clues:

If this is the same site, and I think it is, it appears that the rockslope had been shotcreted and that the tunnel portal had been extended out to provide rockfall protection. I would speculate that the collapse was larger than expected.

Fatality-inducing slope failures on hydroelectric projects in the Himalayas

Unfortunately, the Upper Madi landslide is just the latest of a long series of landslides associated with hydroelectric projects in the Himalayas.  I documented these in a paper and presentation at the Vajont 2013 conference last October in ItalyThe paper can be downloaded for free from the conference website.  In that paper I noted that in recent years there had been 500 deaths in 37 landslide events associated with hydroelectric projects, nearly all of which had occurred in the Himalayas.  In addition, on 27th July a landslide associated with the Xiluodu Dam in Yunnan Province in China killed 12 people and injured three.

The incidence of fatality-inducing landslides for hydroelectric schemes in the Himalayas is far too high, suggesting that better landslide hazard management is needed urgently.

Comments/Trackbacks (0)>>

24 April 2014

Landslide losses – raising the profile in the USA

Landslide losses in the USA

The two dramatic landslides of the last few weeks in the USA has undoubtedly raised the profile of this natural hazard.  The Oso landslide in Washington State remains high on the news agenda, and was even visited by the President earlier this week.  Meanwhile, the fascinating East Gros Ventre Butte landslide in Jackson Hole continues to creep, with devastating impacts on the family who owned the house at the crest of the slide.

These two events have suddenly driven a new level of interest in landslide losses in the USA, which may be a good outcome from these tragic events.  National Geographic has produced a fantastic graphic illustrating the distribution of landslides across the USA since 2007:

This graphic uses data that Dalia Kirschbaum and colleagues at NASA have been collecting on landslide events around the world.  It shows reported landslides in the USA since 2007.  Note that the National Geographic page has more detailed explanations of the distribution of the observed landslides, so I won’t attempt to repeat it here.  However, in general reported landslide incidence is highest where there is a combination of steep terrain, high rainfall, active tectonic processes and people.  Those four factors alone provide an explanation for most (but not all) of the landslides shown above.

In addition, there has been increased interest in State and more local-level landslide inventories.  So, for example, this article describes the incidence of landslides in Oregon, and includes this map of landslides across the state:

landslide losses

KOIN6 and Oregon Department of Geology and Mineral Industries

A key issue in the US media is why there is no national scale landslide mapping programme in the USA – a good question!  In my view landslide work in the US has been grossly underfunded compared with other hazards (this article notes that the entire landslide programme at Federal Level costs $3.5 million per year – in comparison the Bingham Canyon landslide alone is thought to have cost about $1 billion), although the landslide teams at the USGS and elsewhere are world class.  In addition, there is little doubt that excellent and comparatively low cost landslide programmes at the State level have been killed off for political purposes.  A subsidiary question is whether warning systems in the most dangerous places could save lives.  The answer is of course yes – principles and techniques of landslide warning systems are well-established and have proven to be effective elsewhere.  They are however expensive and labour-intensive.  Unfortunately, both the Jackson Hole and Oso landslides will soon slip from the memory of most, and I fear that little will change.  Expect further hand-wringing when the next large event occurs – and given that we appear to be entering El Nino conditions, that may not be too far in the future.

Landslides are probably the most preventable of all natural disasters.  At a national level there is no excuse for failing to manage them properly.

Comments/Trackbacks (3) >>

18 April 2014

Pathe news – historic landslide films

Pathe News

Pathe NewsBritish Pathe News was a company that generated short (four minute) films summarising the latest news, which were then shown in cinemas in the UK and elsewhere.  The company was active from 1910 to 1970, before being killed off by television news coverage.  The archive is clearly a wonderful source of information about world events over this period.  In the last few days, British Pathe have uploaded about 88,000 films onto a Youtube Channel.  The films are genuinely amazing – I thoroughly recommend a browse if you have an hour or two.

Pathe News and landslides

Unsurprisingly, situated within the archive are various films showing landslides.  I will highlight just a small number here:

There is some spectacular footage of the aftermath of a rockslide-induced tsunami in Norway in 1936, which the film indicates killed over 70 people:


There is also a nice piece of footage of the aftermath of a landslide on the Adriatic Coast in Italy in 1956, which knocked a steam train into the sea:


There is some quite dramatic footage of the aftermath of a large rockslide in Los Angeles, USA in 1937:


And finally there is some desperately sad footage of the aftermath of the 1966 Aberfan landslide disaster in Wales:

Comments/Trackbacks (0)>>

16 April 2014

Arroumd – an interesting rock avalanche in Morocco

The Arroumd Rock Avalanche

An interesting paper has just been published in the Bulletin of the GSA by Philip Hughes and colleagues (Hughes et al. 2014) on the Arroumd Rock Avalanche in the Atlas Mountains of Morocco.  This is a very interesting feature, very clearly visible on Google Earth:



The village labelled Aremd, which seems to be more commonly known as Arroumd, is built on the toe of the landslide deposit.  The landslide itself have originated from the steep slopes in the background, traveled down the side valley and been deposited in the valley floor and in part in the main valley.  The resulting deposit has blocked the main valley to a degree. The image below (from here) shows the landslide deposit and the village from upstream – there appears to be the scar from a breach event on the left side; the boulders that form the avalanche are clearly visible to the right of the town:


This deposit has been discussed by geologists for over 130 years, with most interpretations indicating that this debris was the moraine from an ancient glacier (very often rock avalanche deposits have been interpreted as having a glacial origin).  More recently, Hughes et al. (2011) first suggested that the deposit might have a landslide origin; this study involved detailed mapping to investigate whether that hypothesis might be correct.

A complex geomorphic history at Arroumd

The conclusion from the mapping and dating of the deposits is that this is an area with a very complex geomorphic history.  Three moraines are present in the valley, but also present are two course, boulder-dominated deposits that are interpreted as having a rock avalanche origin.  This suggests at least two major landslide events, which have in turn modified the older moraines.  These two major rock avalanche deposits have both been dated to 4,500 years BP (+/-500 years), suggesting that they occurred quite close together in time (which is of course quite common for large rockslope failures).

Unfortunately, there is no simple geomorphic indicator of the likely trigger of rock avalanches, and many such failures occur without a trigger.  Thus, any discussion of a likely trigger event for the Arroumd Rock Avalanche is speculative.  In this case, Hughes et al. (2014) suggest that the proximity of the nearby, active Tizi n’Test fault may well be the culprit.  I think that it would now be really interesting and worthwhile to trench the fault to see if there is a movement event with a similar date.

Finally, Hughes et al. (2014) point out that Arroumd is not the only valley-blocking rock avalanche in this part of the Atlas Mountains – indeed there is an even more spectacular example to the south at Lac d’Ifni, clearly evident on Google Earth:



In this case the landslide has blocked the valley and has not breached, allowing a lake to form.  This lake is now mostly filled with sediments, leaving just the small remnant Lac d’Ifni behind.


Hughes, P.D., Fenton, C.R., and Gibbard, P.L., 2011, Quaternary glaciations of the Atlas Mountains, North Africa, in Ehlers, J., Gibbard, P.L., and Hughes, P.D., eds., Quaternary Glaciations—Extent and Chronology, Part IV: A Closer Look: Amsterdam, Elsevier, p. 1071–1080.

Hughes, P.D., Fink, D., Fletcher, W.J. and Hannah, G. 2014. Catastrophic rock avalanches in a glaciated valley of the High Atlas, Morocco: 10Be exposure ages reveal a 4.5 ka seismic event. Geological Society of America Bulletin, doi: 10.1130/B30894.1

Comments/Trackbacks (0)>>

14 April 2014

Dart River (Te Horo) landslide complex in New Zealand

The Dart River (Te Horo) landslide

The Te Horo landslide on the Dart River in New Zealand is the subject of a new report (Cox et al. 2014) that has been prepared by GNS Science that has been made available online.  The landslide is fascinating and unusual.  It consists of a very large (56 million cubic metre, 0.9 square kilometre) compound rock and debris slide on a steep mountain side.  Below the landslide is an enormous debris fan – this has an estimated volume of 100 million cubic metres – which has diverted the Dart River in the main valley below.  This image, taken from Cox et al. (2014) provides a panoramic view of the landslide:


The slide caused some problems earlier this year when a period of more intense activity in the landslide led to increased sediment deposition on the debris fan. The volume of material entering the channel was sufficiently large that the fan blocked the river, allowing a large lake to form upstream:


Unusually, this type of valley blocking landslide is likely to pose a low level of hazard.  Despite the large lake volume (estimated at 11-15 million cubic metres), the low gradient of the blockage and the presence of large boulders in the deposit means that a rapid breach event is unlikely.  Thus, the lake is likely to persevere for some considerable time and may fill with sediment from upstream.

Interestingly, the landslide itself appears to be evolving quite quickly.  This is a sketch cross section of the landslide from Cox et al. (2014):


The report notes that the area of activity of the landslide itself has moved from an elevation of 800 – 900 m in 2010-12; 900 – 1200 m in 2013; and 1000 – 1350 m in 2014.

There is a great more detail about the landslide, the debris fan and the lake, and many interesting images, in the report, so do take a look.


Cox, S.C., McSaveney, M.J., Rattenbury, M.S., and Hamling, I.J. 2014. Activity of the landslide Te Horo and Te Koroka Fan, Dart River, New Zealand during January 2014. GNS Science Report 2014/07. 45 pp.

Comments/Trackbacks (2) >>

7 April 2014

Two new landslide videos: Hannover Point, UK and Jalisco, Mexico

A large landslide in Jalisco, Mexico

Liveleak has this new video of a large landslide that apparently occurred in Jalisco in Mexico over the weekend.  I know no more details, but the impatient driver of the red car was lucky that the landslide did not occur 20 seconds later:


A small landslide at Hannover Point

Meanwhile there is a much better quality video of a somewhat less dramatic landslide at Hannover Point on the Isle of Wight, this time on Youtube:


The On The Wight website has a good description of how this video was collected.

Thanks also to Dr Phil Collins of Brunel University (@PhilCollins_UK) who tweeted that the start of this landslide event is available separately on Youtube:


Comments/Trackbacks (2) >>

2 April 2014

The Oso (Steelhead) landslide in Washington State – could it have been foreseen?

The Oso (Steelhead) landslide in Washington

The best news about the Washington landslide is that, as forecast, the number of victims is likely to be rather lower than was feared at one stage (I noted earlier that this might well be the case).  The official toll has now reached 28 people known to have been killed, with a further 22 remaining missing.  To have recovered 28 bodies to date is an impressive achievement, and this will have been deeply traumatic for those involved in the search and rescue operations.

So now attention, in the media at least, is focused on whether the landslide could have been foreseen.  Opinion on the internet varies greatly (no surprises there!), ranging from the view that the risk was tolerably low to the view that it was completely foreseeable.  I’ll nail my colours to the mast – to my mind this was foreseeable event, and as such the disaster represents a failure of hazard management.

Using the past as a guide to the future

The simplest way to understand the likely future behaviour of a landslide is to examine how similar landslides have behaved in the past.  In the case of the Steelhead landslide, this was very simple.  Immediately to the west of the landslide there is another large failure.  This landslide is very obvious on the LIDAR imagery.  Dan McShane provided an analysis of the Steelhead landslide history from Google Earth images within a day or so of the slide event, and included LIDAR imagery (from before the most recent movement of the Steelhead landslide) that showed this other failure to the west:


In this image the Steelhead landslide is highlighted on the right, and this much larger landslide is highlighted in the centre of the image.  Note that there is another landslide scar between these two landslides.  The key aspect of this larger landslide is the extensive area that the debris has covered.  And look at the morphology of the deposit as shown in the imagery, and in particular the structure of the material that has flowed onto the valley floor, as shown in this BBC/AP image:


So this bluff has a long history of landslides, and the evidence on the ground suggests that they can be very mobile.  Note also from the LIDAR imagery the shape of the land above the old Steelhead landslide scar – this promontory does not look like a stable configuration in such a weak deposit.

The defence of the emergency planners (as reported by Andrew Alden) in Washington seems to be that after the 2001 movement event at Steelhead the landslide was mitigated by protecting the toe from erosion by the river.  There is little doubt that a large amount of work was undertaken in this respect.  This is a Google Earth image of the foot of the landslide in 2003:


And this is the same view in 2007:



Clearly the river was relocated and the toe of the landslide was protected.  Note the houses that are visible in the second image that were not there in 2003.  However, for these very large landslides it is likely that the failure is driven by groundwater driven processes in the upper part of the slope.  Protecting the toe will have done little to prevent this from occurring.  Thus, the mitigation did not address the primary concern.

The 2001 landslide left material high on the hillside that was sitting above a scar that was far too steep (see the image in this post).  The LIDAR data suggests that the runout from such a collapse could be extensive.  In that context I find the decision to build new houses at the foot of the landslide to be very surprising.

Comments/Trackbacks (6) >>

28 March 2014

The Oso (Steelhead) landslide: mechanisms of movement and the challenges of recovering the victims

The Oso (Steelhead) landslide – mechanisms of movement

In the last three days the desperate search for the up to 90 missing people at Oso has continued (see my earlier post on the landslide). Survival rates for landslide victims are very short, so this is not a rescue operation any longer. During this time some very interesting information has emerged about the landslide in the form of a seismic record of the slide.  There is an excellent blog post from Kate Allstadt about this seismic signal on the PNSN blog – their understanding of this data is much better than is mine, so I won’t try to replicate it here.  For me the most interesting aspect is the double seismic signal generated by the landslide, which indicates two major movement phases (followed by lots of small slips, which we would expect):


Kate Allstadt via the PNSN blog


The start of the two movement events were about 4.5 minutes apart; the first lasted about 2.5 minutes, the second was somewhat shorter and less energetic (i.e. the movement rate was probably slower).

So what does this tell us about the landslide?  We need to compare this with an image of the landslide after the failure – this Wikipedia image remains the best that I have seen for getting an overview of the whole landslide:




For comparison, I have tried to create a pre-failure view from Google Earth – this uses a Landsat image from last July:



To me the most logical explanation that ties the images and the seismic data together is a two phase movement event, as Kate Allstadt suggested:

The first movement event would have been a very rapid and violent collapse of a lower landslide block.  I have labelled the remains of this block, now mostly broken up, on the image below:



This first collapse event would have generated a very rapid mudflow as it collapsed onto the sediments below – this may have caused most of the damage.  The best video I can think of to illustrate this initial collapse event is the Po Selim tin mine landslide in Malaysia – although note that in the case of the tin mine the mudflow may have been more energetic due to the possible presence of pools of water in the floor of the quarry:

The violence of the process does, fortunately, suggest that the victims will not have suffered for long.

Recovering the victims

To date 25 victims have been recovered.  This might seem low when there may be as many as a further 90 people buried in the debris.  However, recovering victims from a mudflow is exceptionally difficult as the remains are likely to be deeply entombed in a material that has very low permeability.  For this reason, the authorities may need to decide to preserve the site as a grave.  If the search is to be continued, a very detailed mapping exercise will be needed.  This will require three key elements:

  1. Attempts will be needed to identify where on the ground each victim was located when the landslide struck.  Were they in their house, in a car on the road, etc.  The better the initial location information the better the chances of finding the final resting place;
  2. The landslide mechanisms will need to be identified in great detail.  Detailed mapping of the landslide will be needed to do this – the investigators will need to ascertain whether victims were pushed ahead of the slide, were entrained within it or were buried in situ.  In my experience the latter is the least likely in this type of slide, but the patterns will vary across the landslide;
  3. Detailed mapping of the human debris will be needed.  The remains of each structure will need to be mapped out.  This will give key information about the dynamics of the landslide at each point, and of course it is also likely that many of the victims would have been close to, or within, buildings.

This is a time-consuming, challenging and sometimes dangerous task, and one that will not necessarily be successful.  Generally speaking this level of work has rarely been undertaken, so it would push the boundaries of our knowledge.  The authorities will need to balance competing pressures in determining a way to proceed.  This may not be an easy, or a popular, choice.

Comments/Trackbacks (1)>>

25 March 2014

The Steelhead landslide in Oso, Washington State

The Steelhead landslide

The death toll from the Steelhead landslide near to Oso in Washington State is continuing to rise.  Latest reports suggest that there are now 14 known fatalities, but 176 people are reported to be missing.  It is quite normal in this sort of event for the number of reported missing people to exceed substantially the actual number of victims, so this maximum toll may reduce in the next few days.  However, it is still likely to be the costliest landslide in terms of lives lost for many years in the USA.

Details are slowly emerging of the landslide history of this site.  It is clear that major landslides have occurred here on many previous occasions; indeed so much so that the landslide is known as either the Hazel landslide or the Steelhead landslide; at this stage I am opting for the matter given that the inundated area is known as Steelhead Drive.

Better images of the Steelhead landslide

I very much appreciate the help that numerous people have given me over the last few days to put together this post – too many to name, but thanks to you all.  The best graphic that gives an overview of the slide is in the Seattle Times:


The original version has a very impressive slider function that allows the user to flip from one image to the other.  I can’t replicate that, but putting the two images side-by-side shows the extent of the devastation.  The number of inundated houses is large, suggesting that the loss of life will be high, especially bearing in mind that the slide occurred on a Saturday.

The best set of aerial images of the slide are on the Flicker page of Governor Jay Inslee – there are some wonderful images there.  This image shows the source of the landslide:


Whilst this image from the same source shows the entire landslide mass.


The landslide has been widely reported as a mudslide.  In terms of the lower portion, which did the damage, this is correct, although in places it might have been more of a mudflow than a mudslide.  However, the upper portion is a rotational landslide – the rotated block with the fallen trees is very clear.  A working hypothesis would be that this block failed catastrophically, transferring load onto the block below, which in turn generated very high pore water pressures, causing fluidisation and a very rapid mudflow that struck the settlements across the river.

The history of the Steelhead landslide

The Yakima Herald has a very nice article that details the chronology of events on the Steelhead landslide.  This includes:

  • 1949: A large landslide (1000 feet long and 2600 feet wide) affected the river bank
  • 1951: Another large failure of the slope; the river was partially blocked
  • 1967: Seattle Times published an article that referred to this site as “Slide Hill”
  • 1997 report, by Daniel Miller, for the Washington Department of Ecology and the Tualialip Tribes
  • 1999: US Army Corps of Engineers report by Daniel and Lynne Rodgers Miller that warned of “the potential for a large catastrophic failure”
  • 25 January 2006: large movement of the Steelhead landslide blocked the river

There is a good presentation about the 2006 landslide available online (NB pdf). This includes the following (somewhat blurry) image of the source of the 2006 landslide:

Steelhead landslide


The slope that formed the scarp of the 2008 slide was undoubtedly very over-steepened, and of course was formed from weak materials.  This looked like an accident waiting to happen.

Comments/Trackbacks (5) >>