8 November 2015
The Samarco tailings dam disaster in Bento Rodriguez
Latest reports from Minais Gerais state is Brazil suggest that there are now two known fatalities from the Samarco tailings dam disaster that devastated Bento Rodriguez, but that a further 28 people remain missing. News reports suggest that about 60 million cubic metres of sludge were released when two tailings dams collapsed. Samarco is now coming in for some criticism for their response to the disaster, and it will be very interesting to see how they handle the inevitable investigation into the dam itself. The fact is that tailings dams should not fail like this – although they do so alarmingly often.
I have seen few images of the location of the source accident itself. The reports consistently suggest that two dams failed. Based on the configuration of the tailings dams, this is most likely to have been the set of dams shown in the Google Earth image below, in which two structures lie upstream of a smaller downstream dam:
So presumably one of the major structures upstream ruptured, triggering a collapse that then over-ran the smaller downstream dam. The best image I have seen that indicates what might have happened in this one (via Estadao):
This image shows the two upstream tailings dams. The left hand dam (the one to the west in the Google Earth image) above is clearly intact, whilst the one to the west appears to have failed. So, based upon this, the problem seems to lie with this dam:
It is hard to know what might have happened beyond this, but the surge that came down the valley was enormous, based on the erosion in the valley below. Note the height of the trimline in the vicinity of the second dam. And of course there is now a vast amount of tailings that are not being retained by any structure. Dealing with this debris must be a priority.
Interestingly there is some information on the internet that there has been recent work to increase the height of at least one of the dams. There is nothing to suggest at this stage that these works had any role in the accident, but it will be a factor in the investigation
6 November 2015
There are reports this morning of a disastrous tailings dam failure affecting the town of Bento Rodrigues in Brazil. Current reports suggest that at least 15 people have been killed and as many as 45 more may be missing.
The mudslides that struck the town appear to have been generated by the collapse of the tailings dam associated with the Germano mine, owned by Samarco, which in turn is a joint venture between Vale of Brazil and BHP of Australia. This is a Google Earth image of the setting, although note that this is a couple of years old:
The best gallery that I have seen is on the uol.com.br website, from which these images are taken. The damage looks near apocalyptic:-
In many ways this event is reminiscent of the dreadful tailings dam failure at Ajkai Timfoldgyar in Hungary in 2010, which I covered extensively at the time. Back in 2009 I wrote about an academic study that had been published in a conference in Alberta that suggested that the frequency of tailings dam failures increases when commodities prices fall. The peak in tailings dam failures occurs about two years after the peak in commodities prices. I wrote:
The relationship between the peak in prices an the peak in accidents is ascribed by the authors to:
The rush to mine quickly means that design and construction standards may be low;
Rapid turn-over of key staff as new (presumably lucrative) opportunities arise during the boom;
The boom drives the development of resources in areas that are known to be difficult;
after the boom there are pressures to cut costs as commodity prices decline;
The boom drives the use of inappropriate designs imported from other locations;
There may be a lack of independent review, presumably to avoid the time delays and costs associated with this.
The prices of commodities are currently very low. Iron ore commodities prices peaked in 2011. It would be interesting to see images of the dam itself.
4 November 2015
Hillslope preconditioning is the term we use to describe the possibility that the behaviour of a slope during an event might also be influenced by the legacy of previous trigger events. This idea arises from a damage mechanics approach to landslides, in which the slope evolves to failure through the accumulation of defects – damage – with time. Hence, each successive trigger event causes the slope to degrade until it is ready – preconditioned – to collapse. There are different ways to conceptualise the damage process – for example, it could be that damage occurs through the progressive development of a sliding surface, such that failure occurs in the trigger event that causes it to become complete. Alternatively, damage might accumulate as a degradation and general weakening of the entire rock mass.
The damage approach to landslide failure is appealing in many ways, not least because it fits with observations of slopes prior to failure, which often show large amounts of deformation prior to collapse. This is one of my favourite examples, from Malaysia. Note the extensive damage to the main (rock) slope on the left side of the image:-
A closer look shows the massive displacement on the shear surface and the damage to the rock mass that forms the landslide:
These images were taken in 2006. The slope has yet to collapse.
Whilst hillslope preconditioning is a really interesting idea, there is little quantitative analysis to ascertain whether it is a real factor in slope behaviour. In a paper just published in Earth Surface Dynamics (Parker et al. 2015) – this is an open access paper, so you can download it for free – we have tried to explore this effect by looking at the effects of two earthquakes in New Zealand. These earthquakes occurred in 1929 – the so-called Buller Earthquake – and in 1968 – the Inangahua earthquake. Both triggered large numbers of landslides, which were subsequently mapped, primarily by my good friend Graham Hancox at GNS Science in New Zealand. The beautiful thing about these two events is that they occurred reasonably closely together (the epicentres were 21 km apart), such that there was an area that was affected by both earthquake events, as this image from the map shows:-
This work was led by my then PhD student Robert Parker, who hypothesised that if the hillslope preconditioning idea is correct then in this overlap zone it might be expected that the probability of landslides being triggered by the second – 1968 – earthquake would be higher than in the rest of the area affected by this seismic event. In other words, we think that the 1929 earthquake might have preconditioned the slopes in the overlap zone, making them more susceptible to failure. This should be characterised by an unexpectedly high occurrence of landslides in this zone compared with the remainder of the area affected by the 1968 earthquake.
The idea is simple, but investigating this hillslope preconditioning effect is actually very challenging. We had to use a complex statistical approach – I won’t describe it in detail here, but you can read about it in the paper. Essentially we had to model the effects of time independent factors – slope angle, lithology, etc – and then correct for these factors in the overlap zone. The analysis suggests that once these factors have been controlled the probability of failure in the overlap zone is higher than would be expected, suggesting that another factor – hillslope preconditioning – is having an influence.
In the paper we emphasise that these results are tentative, and that more work is needed. In particular, the dynamics of the interactions between slopes and seismic waves are so complex that it could be that there is another factor, for which we have not controlled, having an influence. Indeed it could be that the results are just random effects. However, the result does seem to be robust and fits with our understanding of slope behaviour.
If this result is correct then it potentially makes the analysis of the potential for slope failure in a future earthquake even more difficult. On the other hand, it might also open up new possibilities for analysing slope response to seismic events, as well as changing the way that we conceptualise the development of failure.
Parker, R.N,., Hancox, G.T., Petley, D.N., Massey, C.I., Densmore, A.L., Rosser, N.J. 2015. Spatial distributions of earthquake-induced landslides and hillslope preconditioning in the northwest South Island, New Zealand. Earth Surface Dynamics, 3 (4), 501-525.
2 November 2015
New landslide videos
In the last couple of weeks some new landslide videos have appeared:
A huge rockslide in Taxila, Pakistan
This one is pretty spectacular, although I have very little information about it:
I would be interested in more information. I don’t think I’ve seen it before, but am unsure. It is truly spectacular!
A rockslide in Mexico
This one was covered quite extensively in the media:
At least 15 cars were damaged as six thousand cubic metres of earth crashed onto the road below. Amazingly, the landslide did not bring down two tower blocks and a private house built precariously near the cliff edge as the ground collapsed underneath them. But shocked drivers and pedestrians below meanwhile tried to flee the debris, and the huge dust cloud that followed. The dramatic incident happened in Santa Fe, one of the major business districts of Mexico City, located in the west part of the Mexican capital city. Parts of the video were taken from the other side of the valley, on dashboard cameras and mobile phones, and they were all uploaded onto social media where they quickly went viral. Officials say people living in the apartment buildings were told there was a risk of landslide back in 2007, but ignored their warnings. The authorities of Cuajimalpa, one of the 16 boroughs of Mexico City to which Santa Fe belongs, also said they had previously advised the inhabitants to abandon the buildings fearing they would fall if there was a landslide.
A rock topple in Switzerland
This pair of landslide videos showing a spectacular rock topple in Switzerland was also extensively covered:
This is the incredible moment a huge piece of rock broke free from the side of a mountain and plummeted 1,000 feet in Switzerland. The incredible footage was captured by a geologist who just happened to be filming when the rock fall occurred on a mountain in the Swiss Alps. According to local reports the 2,000 cubic metres of solid rock that fell caused an avalanche on the mountain of Mel de la Niva, near Evolene. Geologists studying the mountain had noted geological activity prior to the incident and the site had been under surveillance since 2013, reported Le Nouvelliste. Major movements were detected last week and the hamlet of Arbey and a road below the mountain had been evacuated and shut the day before.
30 October 2015
The Big Horn Mountains landslide
There is quite a lot of interest in the media in a landslide that has developed in the last fortnight in the Big Horn Mountains of Wyoming. This was first reported by SNS Outfitter and Guides on their Facebook site, with the following impressive picture:-
The caption was:
This giant crack in the earth appeared in the last two weeks on a ranch we hunt in the Bighorn Mountains. Everyone here is calling it “the gash”. It’s a really incredible sight.
A couple of days later they followed this up with a further image:-
The caption this time was:
Since so many people have commented and asked questions, we wanted to post an update with a little more information. An engineer from Riverton, WY came out to shed a little light on this giant crack in the earth. Apparently, a wet spring lubricated across a cap rock. Then, a small spring on either side caused the bottom to slide out. He estimated 15 to 20 million yards of movement. By range finder, an estimate is 750 yards long and about 50 yards wide.
29 October 2015
Tonzang landslide event in Burma
In late July and early August extreme rainfall in the Tonzang area of Burma triggered extensive landslides and floods. Whilst there were some media reports of these events (and a good blog site has reported upon them), and a few weeks ago I posted some images, details of this event have remained sketchy. This event has been particularly intriguing as Goran Ekstrom and Colin Stark from Columbia University picked up two landslide seismic events, and of course these are only generated by the largest landslides. Unfortunately the summer is monsoonal in Burma, meaning that it has not been possible until now to obtain good quality satellite imagery. In the last few days NASA kindly tasked the EO-1 satellite to image the area, and Colin has equally kindly provided some images and information.
What is now clear is that this was a most unusual landslide event. The satellite image below shows and overview of the landslides triggered by the rainfall. Note that this image does not cover the entire area, so there are more landslides to be found as yet. The brown areas in the mountains are the multiple landslides triggered by the rainfall – note just how many there are:-
I have posted before on the largest of the landslides, which is clearly visible in the image above, based upon the UNOSAT mapping of a part of the slide. We now have decent imagery of this event – and it is truly enormous:-
It is hard to get a scale on this from the image, but from the crown of the failure to the toe it is 5.9 km. This is a genuinely remarkable event – without doubt one of the largest rainfall induced landslides in recent years. It is clear that the landslide has blocked the valley and a lake has developed upstream, although this is not very large. Note the many other landslides visible on the image, also triggered by this rainfall event. These are very diverse, including channelised debris flows and translational failures.
However, the two seismic events that Goran Ekstrom detected do not appear to correspond to this landslide. These are the two velocity / trajectory graphs for those two slides, generated through the inversion of the seismic data:-
These are both large landslides, but neither the trajectory nor the runout distance correspond to the large event shown above. The only potential candidates on the imagery lie in this area of landslides to the south of the largest event:-
It is not clear to me that these landslides are the correct candidates, which may mean that there are further large landslides that have yet to be captured on the satellite imagery. There is clearly more work to be done, and it will be very interesting to see what emerges when the rest of the area is captured on satellite imagery in the coming weeks.
Many thanks to Colin Stark and Goran Ekstrom and Columbia University and Stuart Frye at NASA for their help with this post.
28 October 2015
The M=7.5 Afghanistan earthquake on Monday was deep (the USGS estimates 213 km), meaning that it affected a large area of mountainous terrain. This is a zone that is highly landslide prone in even static conditions, so landslides were an inevitable consequence. In the aftermath a number of videos and images of landslides have appeared in various places.
Videos of landslides from the Afghanistan earthquake
This video shows a comparatively minor but visually dramatic terrace collapse in the Gilgit region of Pakistan:
This video appears to be a somewhat larger landslide, although I suspect that the nature of the material makes it appear worse than reality. It is certainly a very dramatic video:
This landslide is reported to have occurred in the Hooper Valley of Pakistan, again in weak materials:
Images of landslides from the Afghanistan earthquake
Meanwhile, there are also various images of landslides triggered by the earthquake. The ever-wonderful Pamir Times for example has this image of rockfall debris in the Gilgit Baltistan area:
Whilst landslide losses are a little unclear at the moment, the Pamir Times also reports four people killed in a rockfall on the Karakoram Highway. The level of damage to the road was clearly high, as depicted in this amazing gallery of images:
The Chief of the National Disaster Management Agency in Pakistan is warning of the potential for further landslides in this area. This is good advice. The Daily Times reports that four members of rescue team were killed, and another 12 injured, in a landslide in the Kalam region of Pakistan yesterday.
It is worth noting that both NW Afghanistan and N Pakistan are hotspots for very large landslides in the uninhabited high mountain regions, many of which are triggered by earthquakes. It will be interesting to see whether any such events have occurred but have as yet not been observed.
23 October 2015
A landslide in an artisanal mine in Colombia
An amazing new video has appeared on youtube showing a landslide in what appears to be an artisanal (i.e. small scale, unofficial) mine:-
The text on the original Youtube posting said:-
Unemployment forces these people to risk their lives every day in these mines. Just last Wednesday there was a large landslide that caused a great tragedy in the village of San Antonio in the municipality of Santander de Quilichao, in an illegal mine where gold is mined . This is due in large part to the neglect of the state, because the rulers are more occupied in their political confrontations in their own town
The video is dated 31st March 2015, which would mean that the landslide occurred on 25th March. But actually it seems more likely that this is the collapse that occurred on 20th April 2014. The reported death toll of this event was 16 people.
The truly horrifying danger of working in these conditions is all too clear, with the miners actively undermining the slope, in deeply weathered soil, in order to access the gold. The failure of the slope itself is rapid and catastrophic, and the rate at which the debris overruns everything in its path is remarkable:
The power of the displacement wave is also dreadful to behold.
20 October 2015
The Mount Steele rock avalanche
Further analysis by Colin Stark and Goran Ekstrom has yielded more detail on the Mount Steele rock avalanche in the Yukon of Canada last week. This zoomed in perspective view of the satellite image provides greater detail:-
The analysis now suggests a mass of the Mount Steele rock avalanche of 45 to 50 million tonnes, which would equate to about 20 million cubic metres, with a runout time of about 110 seconds. This is the velocity profile from the seismic data – note the comparatively rapid acceleration phase and a more leisurely deceleration:
Whilst this is a chart of the track of the landslide with the acceleration vectors overlain. The landslide starts in the bottom left corner on this plot:
And finally, this is the long profile of the landslide track from the seismic data. The arrows indicate the velocity of the landslide at each point.
Interestingly, this is second giant landslide in the St Elias area this year alone. On 11th August there was another event that dropped about 40 million tonnes of rock onto the Turner Glacier, not far away from this event. In addition, in July 2007 there was another large rock avalanche from the flanks of Mount Steele. This landslide, which is reported in Lipovsky et al. (2008) [note this paper is available online here], was even larger, with a volume of about 108 million cubic metres. This is an image of the landslide from the paper:
Lipovsky, P.S., Evans, S.G., Clague, J.J.., Hopkinson, C., Couture, R., Bobrowsky, P., Ekström, G., Demuth, M.N., Delaney, K.B., Roberts, N.J., Clarke G., and Schaeffer, A. 2008. The July 2007 rock and ice avalanches at Mount Steele, St. Elias Mountains, Yukon, Canada. Landslides. 5 (4):445-455.
19 October 2015
Mount Steele Rock Avalanche
On 12th October at 02:02:32 UT a giant rock avalanche occurred on the flanks of Mount Steele in the Yukon area of Northwest Canada. The landslide was detected by Colin Stark and Goram Ekstrom of Columbia University using the global seismic network, and I am posting this news with their kind permission. Since the landslide occurred the availability of cloud-free satellite imagery has allowed the location to be identified.
The landslide is large – I will give the full details from the seismic inversion tomorrow – but Stark and Ekstrom estimate a mass of about 60 million tonnes, which yields a volume of about 24 million cubic metres. The crown of the landslide is at an elevation of about 3350 m, the total vertical distance is about 2200 m to the toe, with a runout over the Steele Glacier of about 3700 m.
Colin Stark has provided this Landsat 8 image of the landslide, draped into a digital elevation model. As can be seen below the landslide detached from the flank of Steele SE, a sub-peak of Mount Steele:
This is an image of the same location on Mount Steele before the landslide, also from Landsat 8 via Colin Stark:
Note the location of the landslide as indicated through analysis of the Landslide Force History (LFH) seismic data. This was within 7 km of the actual location of the Mount Steele rock avalanche, which is impressively close. The seismic data suggests that this rock avalanche had a peak velocity of about 60 m / sec, which is about 220 km/h. An interesting aspect of this landslide is the smooth morphology of the slope that failed compared to those around it, as seen in the above image. I am unsure as to whether this is significant.
I will provide more detail about, and further images showing, this landslide tomorrow.