25 September 2019
The Mirpur earthquake in Pakistan: images of lateral spreading
On 24th September 2019 an M=5.6 earthquake struck Mirpur in NW Pakistan. Whilst the Mirpur earthquake was comparatively small, it was also shallow, meaning that a significant area will have suffered high peak ground accelerations. The USGS has generated a map of earthquake intensity (the contours on the map), enhanced by the shading, which shows areas of liquefaction potential:-
At the time of writing, over 30 people have been reported to have been killed, whilst at least 450 people were injured. Interestingly, the area with the highest intensity of shaking coincides with both the banks of the major Jhelum River and the margins of part of the huge Mangla reservoir:-
Thus, whilst this earthquake is not large, it has the potential to generate both liquefaction and lateral spreading. Early images coming out of Pakistan show that the earthquake has had just this effect. For example, this image (from the Independent), clearly shows lateral spreading on the banks of the river. Note that the material appears to be made ground:-
Whilst this image, also from the Independent, also appears to show this area of lateral spreading:-
As usual, it is difficult at the moment to ascertain the full extent of these types of impacts – this should become apparent in the days ahead. It will also be interesting to find out the proportion of the building collapses that are associated with these lateral spread events, rather than simple structural failure under high peak ground accelerations.
24 September 2019
A new analysis of the deadly Anak Krakatau flank collapse
Almost a year ago, the collapse of the flank of Anak Krakatau in Indonesia generated a tsunami that killed over 400 people. The flank collapse occurred without warning, although there had been concerns for a long time that sch an event could occur (as is the case for other volcanic flanks of course). The journal Geology has recently published a paper (Williams et al. 2019) that provides a first detailed analysis of this event. The findings are in some ways surprising. I previewed the article when it was published on Earth ArXiv; the full paper has now been published. I should also note that the authors acknowledge my earlier comments in the paper.
Williams et al. (2019) provide some interesting information about the consequences of the event – for example, the tsunami that caused the losses had a maximum height of 1.40 m; it killed 431 people and injured a further 7,200; it destroyed 1,778 houses; and it damaged 434 boats and ships. But the focus of the work is the use of satellite data to reconstruct the chronology of events and, most importantly, to analyse the magnitude and dynamics of the flank collapse itself. The key conclusion about the sequence of events is contained in the schematic diagram to the left.
In essence, the volcano failed in a large rotational landslide that removed the flank of the volcano. The toe of the slide (and thus the main mass of the landslide) was below sea level. This triggered the partial failure of a second portion of the flank of the volcano (see diagram B in the schematic illustration), but this section did not proceed to full failure at this point. Subsequently, the volcano replumbed to generate a new vent through the basal surface of the landslide (see schematic diagram C). The eruption through this vent involved the ingress of sea water, generating a violent phreatomagmatic eruption. This eruptive event removed the remainder of the flank of the volcano, including the partially slipped landslide mass.
The subsequent eruptive events on the volcano then led to the generation of a new, more stable morphology, allowing the activity to stabilise (see schematic diagram D).
The surprising element of this analysis is the scale of the main Anak Krakatau flank collapse. The satellite imagery allows the construction of detailed cross sections, although clearly some assumptions need to be made about the underwater configuration given the lack of bathymetric data. But this analysis yields a volume of about 4 million m³ for the subaerial (i.e. above water) component and 100 million m³ for the below sea level component of the landslide. Whilst this is a very large landslide, it is remarkably small for a flank collapse, and it is also remarkably small for a landslide to have generated such a large tsunami. Williams et al. (2019) compare their findings with a previous study that modelled the generation of a tsunami from an Anak Krakatau flank collapse, but which assumed a failure volume of 280 million m³. The observed waves generated by the actual flank collapse were on a similar scale to those modelled for the much larger event. Interestingly, they also moved through the water much faster than the model suggested. This suggests that the model is under-predicting tsunamis generated by large landslides.
So, overall, this is a really interesting analysis of the Anak Krakatau flank collapse. The results have some profound implications for localised but highly destructive tsunami generation from these events, and the study implies that we will need to look again at the ways in which tsunamis are modelled.
Rebecca Williams, Pete Rowley, Matthew C. Garthwaite. 2019. Reconstructing the Anak Krakatau flank collapse that caused the December 2018 Indonesian tsunami. Geology. https://doi.org/10.1130/G46517.1
23 September 2019
Kerala – the hungry rains
The Indian financial newspaper Mint has published an article last week, The Hungry Rains, which examined the impact of the 2019 monsoon on Kerala in western India. Kerala is the main coastal state lying along the southwest part of the country. The main take home message from the article is that Kerala is enduring an ecological crisis that is making the state exceptionally vulnerable to the effects of the monsoon, resulting in extensive destruction and high levels of economic loss. Kerala has suffered devastating losses two years in a row, with the exemplar being the Kavalappara landslide in August, which killed about 60 people.
The Mint argues that poor land use management lies behind many of the problems in Kerala:
“A journey through Kerala shows it’s not just about the weather. Experts from the Kerala State Disaster Management Authority spent the first week of September studying what led to the mudslide that wiped out Kavalappara. Their findings have yet to be made public but a person privy to them says, on condition of anonymity, that apart from the heavy rains, unscientific cultivation of rubber across the hilltop is to blame. The strength of the soil, its ability to resist deformation and lateral motion, has been destroyed by the rubber estates, he says. In other parts of the state, which saw heavy inundation and deadly landslides, they found a clear correlation between the damage and human activities such as quarrying or encroachment of riverbeds. People know about the changes but are not aware of the dangers.”
The vulnerability of India to landslides is fascinating. Whilst the focus is (correctly) often on the Himalayan Arc, there are also substantial challenges in the west of the country. The map below shows the rainfall-induced landslide susceptibility for India. I have created this map from the NASA landslide susceptibility dataset from Dalia Kirschbaum and Thomas Stanley:-
And this map shows the distribution of fatal landslides in India from 2004 to 2016, as per the work that I have undertaken with Melanie Froude. The cluster of landslides in western India is clear:-
The summer monsoon in India consists of air drawn from the seas to the southwest of the country, meaning that the coastal areas suffer from high levels of rainfall. This is the seasonal (monsoon) rainfall map for 2019, from Monsoon Online:-
Note the very high level of expected rainfall across western India (the “normal” map in the centre), and the exceptional positive anomaly in 2019 (the “departure” map on the right). Thus, Kerala would be expected to have very high levels of rainfall, but this year the total received was significantly higher than average.
The Mint article considers the changes that are needed to try to manage the hazard:
What is the solution? A serious effort must begin with addressing the elephant in the room—how land is governed, says V. Venu, chief executive of the Rebuild Kerala Initiative, a special purpose vehicle floated by the government to rebuild the state after the back-to-back floods and prevent future destruction. “Today’s legal framework practically lets you build pretty much whatever you want, a resort or a house or a commercial building. They have a few stupid restrictions. And, if you fulfil those, for everything else you will get a licence,” he says.
Sadly, at present there is little prospect of meaningful change, so the losses are likely to mount in the years ahead.
20 September 2019
The 1985 Stava tailings dam disaster
I’m currently undertaking some work examining the runout of mine waste failures, during which I’m looking back at some old case studies. This is a follow up to the paper that Melanie Froude and I published last year that suggests that loss of life from slope failures in the mining industry, most particularly where there is poor regulation, is increasing with time. This has led me to look back at one of the worst failures of this type in modern history, the 19th July 1985 Stava Tailings Dam failure in Italy.
The location for the disaster was the village of Stava, in Trento. At the site, two tailings dams, built to contain the waste from a fluorite mine, had been constructed. The site, and the collapse, is described in a post-event analysis (Chandler and Tosatti 1995), available online, whilst there is also a nice reflective piece about the role of regulation (Luino and Degraff 2012), also available online. The latter includes this image of the tailings dams prior to failure:-
The dams collapsed on 19th July 1985 at 12:22. The initial failure occurred in the upper tailings dam, which then induced collapse of the lower facility in a domino effect. At the time of failure, 300,000 m³ of material was stored in the two tailings dams. Of this, 180,000 m³ was released in a single event, which mobilised into a very rapid mudflow. Eyewitness accounts suggest that the rate of movement was sufficiently great to generate an air blast that shredded the trees along the path of the flow. There is a good seismic data for the landslide, which suggests that it reached a peak velocity of 27 metres per second (about 100 km/h or 60 mph). The mudflow struck the houses located directly below the tailings dams before sweeping down to the village of Stava, located about 800 m below the lower dam. Luino and De Graff (2012) report that Stava was struck at 12:25, and the 20 or so buidings were completely destroyed in just 13 seconds.
The flow then travelled down the valley, ultimately sweeping through parts of the town of Tesero, located about 3 km downstream. The before and after image below, also from Luino and De Graff (2012), shows the extraordinarily destructive power of this flow:-
The mudflow from the Stava tailings dam failure finally stopped when it reached the main channel of the Avisio river at Tesero. The seismic data suggests that the movement event arrested just ten minutes after the initial collapse. The village of Stava had been completely destroyed, and there was extensive damage in Tesero. In total 268 people were killed, and 56 houses, six industrial buildings and nine other buildings were destroyed.
Subsequent analysis of the site (see Chandler and Tosatti (1995) suggests that the stability of the tailings dams was unacceptably low, primarily because the underlying ground was poorly drained, the construction meant that the dams lacked adequate drainage (allowing high water pressures to develop, and preventing proper consolidation of the tailings), the ponds were being recharged with runoff from the adjacent drainage basins, and the upper dam was unacceptably steep, with a part of the retaining structure being sited on tailings from the lower pond.
The failure led to legal action against those responsible for the dam. In June 1992 a total of ten individuals were convicted of crimes that included culpable disaster and manslaughter of multiple individuals, and were jailed.
The Stava Foundation has a very detailed website that documents the disaster in order to remember the victims, the aim of which is to promulgate the lessons from the Stava tailings dam failure. Sadly, the industry has yet to heed the lessons adequately.
Chandler, R.J. and Tosatti, G. 1995. The Stava tailings dams failure, Italy, July 1985. Proceedings of the Institution of Civil Engineers Geotechnical Engineering, 113 (2), 67-79.
Luino, F. and De Graff, J.V. 2012. The Stava mudflow of 19 July 1985 (Northern Italy): a disaster that effective regulation might have prevented. Natural Hazards and Earth System Sciences, 12, 1030–1042.
19 September 2019
Mine waste landslides at the Kumtor Goldmine in Kyrgyzstan
In the course of some work that I’m undertaking on human-induced landslides, I have come across an interesting article (Torgoev and Omorov 2014 – available online) about landslides at the Kumtor Goldmine in Kyrgyzstan (location = 41.868, 78.169). This site is unusual as the mining is being undertaken at an unusual elevation, 3,000 – 4,400 metres above sea level, at a site that has active glaciers. The mining waste is being dumped into a glaciated valley, allowing the development of a large landslide. Fortunately this is slow-moving compared with many other mine waste landslide events. The site can be seen in the Google Earth image below, collected in August 2016. I have marked the toe of the landslide:-
Torgoev and Omorov (2014) observe that by 2013 about 1 billion tonnes of mine waste and 100 million tonnes of glacial ice had been removed from the Kumtor Goldmine, and dumped in the valley. Most of the waste was dumped onto active glaciers, primarily the Davydor Glacier. This has triggered movement that is a strange hybrid of glacial flow and a landslide, although as the image above shows, the current morphology is primarily that of a mine waste landslide.
The largest landslide (but note that there are others at the site), has a volume of about 70 million m³, and a total movement distance of about 3 km based on the Google Earth image. Torgoev and Omorov (2014) report that it is about 85 metres thick, and is moving on a surface with a gradient of about 5°.
Fortunately, the rate of movement of these dumps is quite low compared with many other mine waste landslides. Geodetic data presented in the paper indicates movement rates in 2012 of about 1.5 metres per day, although a higher movement rate event was recorded in April 2013, when rates reached 5.3 metres per day. This caused the loss of some buildings:-
The landslide at the Kumtor Goldmine continues to move. This image is a Planet Labs PlanetScope image from 11th August 2019. I have rotated it to allow easier comparison with the Google Earth image above:-
This image suggests that the landslide has continued to advance, probably more slowly than previously, whilst the toe of the landslide has started to spread laterally. The toe of the landslide appears to consist primarily of material entrained into the slide from the slopes below the dump.
This is an unusual and interesting hybrid mine waste landslide. I wonder how its dynamics might change as global heating induces temperature rises (and thus thawing of the ice) in the landslide mass.
Torgoev, I. and Omorov, B. 2014. Mass Movement in the Waste Dump of High-Altitude Kumtor Goldmine (Kyrgyzstan). In Sassa K., Canuti P., Yin Y. (eds) Landslide Science for a Safer Geoenvironment. Springer.
Planet Team (2019). Planet Application Program Interface: In Space for Life on Earth. San Francisco, CA. https://www.planet.com/
18 September 2019
Landslides in Art Part 30: Erica Putis
This is the 30th edition of my (very occasional) series on the ways in which landslides are depicted in various forms of art. Part 29 can be found here.
This time I’m featuring a piece of work by Erica Putis, who describes herself, and her work, thus:
Erica Putis grew up in a small town in Vermont where the grass was green and the winters cold. During this time, she wandered for hours around her family’s 20 acres, continuously in awe of nature and her surroundings. Creating was always something that happened and never stopped even when she grew up…Erica’s work is inspired by her love of nature, science and beauty. She tries to find the balance between the terrifying beauty of mother nature and the fascinating anomalies of the science world. These subjects have always motivated her creativity and given her a sense of awe while living on this little place we call Earth.
This work is from a collection of work entitled Natural Destruction Series, which she describes as follows:
I grew up in Vermont and am a nature girl at heart. I am totally fascinated and terrified of natural destruction. In our current state of affairs due to climate change these beautiful but scary phenomena will continue to increase with frequency and strength. I am continuously amazed at Mother Nature and how we lack control over her. Watching the weather channel, confronting my phobia of tornadoes and studying science and geology in my spare time has inspired this series.
The works depict various geological and meteorological phenomenon, including this painting of a landslide:-
This painting appears to be inspired by this photo of a landslide on Cecil Lake Road in Canada by R. Couture:-
The painting rather beautifully captures the sad destruction of the natural environment caused by the landslide. There is a sense of melancholy and loss, which cannot be captured in a photograph.
17 September 2019
Landslides and air blasts
An interesting aspect of some very large and rapid landslides is that they generate air blasts that can be highly destructive in their own right. A good example is the Langtang landslide, triggered by the 2015 Gorkha earthquake in Nepal. In that case the air blast caused huge damage even in areas not reached by the landslide debris itself. However, there has been comparatively little analysis of this phenomenon to date.
A recent paper (Zhuang et al. 2019) provides an analysis of the air blast generated by one of the large landslides triggered by the 2008 Wenchuan earthquake in China. The landslide in question, the Wenjia rock avalanche (also known as the Wenjiagou landslide), has a volume of about 50 million cubic metres and traveled about 4500 metres:-
The study by Zhuang et al. (2019) is based upon a numerical analysis of the air blast generated by the landslide. The model suggests that the rock avalanche itself lasted 210 seconds and moved with a maximum velocity of 65 metres per second (about 230 km per hour). The authors describe evidence on the ground of the resulting air blast, including mature trees snapped in half:-
This type of damage was found on both sides of the landslide path and downslope from the landslide deposit. The analysis suggests that a blast wave started to form about 30 seconds after the initiation of sliding. The maximum wind speed was 35 metres per second (about 125 km per hour). This is the equivalent of a Force 12 hurricane wind, which is associated with complete devastation. The effects of the air blast extended 750 metres beyond the margins of the landslide. The model suggests that the topography had a strong control on the form of the blast that was generated.
We rarely account for the effects of landslide air blasts in hazard models. This paper is further evidence that for the very largest and most rapid landslides they can be an important mechanism for causing damage.
Zhuang, Y., Xu, Q. & Xing, A. 2019. Numerical investigation of the air blast generated by the Wenjia valley rock avalanche in Mianzhu, Sichuan, China. Landslides. https://doi.org/10.1007/s10346-019-01253-0.
Tang, Y., Zhang, Z., Wang, C. et al. 2018. The deformation analysis of Wenjiagou giant landslide by the distributed scatterer interferometry technique Landslides 15: 347. https://doi.org/10.1007/s10346-017-0917-5.
16 September 2019
An update on landslides from the 2019 Northern Hemisphere monsoon season
At the 2019 Northern Hemisphere summer monsoon season starts to decline, I have run a quick analysis of the landslides that I have recorded to date (following up on my post of a month or so ago looking at the number of landslides in July). Whilst the 2019 total will increase slightly as I work back through the dataset to validate it (a process that usually uncovers some unrecorded events), the graph below is the dataset up to 11th September. The plotted data shows the cumulative number of landslides and the cumulative total number of fatalities worldwide up to that date:-
The effects of the 2019 Northern Hemisphere summer monsoon season are clear, showing a distinctive increase in the rate of fatal landslides from about day 180 (29th June, marked with the first vertical dashed line). This dramatically increased rate persists until about day 230 (18th August), when the number of recorded events reduces back to close to the normal (background) rate. Some of these landslides after this event are of course in the monsoon affected areas – for example, although they are not in the database as yet, Nepal suffered some fatal landslides at the end of last week, with one event killing six people.
Overall, the 2019 landslide year looks to be very similar to 2018. As of the end of August I had recorded 315 fatal landslides in 2019 (this number will rise as noted above), against 319 at the same point in 2018. The average for the period 2003 to 2018 (i.e. the entire range of the dataset) is 266 landslides, so the last couple of years have been notably worse than the long term average.
I would also like to highlight an interesting aspect that I cannot really explain at the moment. This graph shows the number of fatal landslides recorded worldwide in August over the period 2003-2019:-
It shows a generally increasing trend; August 2019 does not look exceptional (but will increase when I improve the data). By comparison, July shows quite a different pattern:-
Again, there is a clear rising trend. But from 2011 onwards the pattern seems to have changed, with the number of recorded landslides become highly variable from year to year. 2019 is clearly the worst year on record.
This apparent change in pattern may or may not be significant, but requires further investigation.
12 September 2019
Landslides during the construction of the Morehall and Broomhead reservoirs in Sheffield
I have previously written about the problems that landslide posed to the construction of the Longendale reservoirs in the vicinity of my home town, Sheffield. My local paper, the Sheffield Star, has an interesting article this week about landslides during the construction of two other reservoirs, Morehall and Broomhead, built to supply water to the city in the early part of the 20th Century. The article, written to a highlight a guided heritage walk to the sites on Saturday 21st September, describes major landslide related construction problems, in particular at Broomhead (location is 53.46, -1.60). Part of the article describes these landslides, with the quotations being provided by Mike Atkinson, a former engineer who carried out work on Broomhead Reservoir during his career (Mike will be leading the walk on 21st September):-
In 1924, workers noticed that the hillside towards Yewtrees Lane, north of the Broomhead site, began slipping as an important overflow channel was being dug. “They realised they needed to do something,” says Mike. “They were digging the channel and the ground kept coming towards them.” A trench was dug in short lengths and a drain up to 50 feet deep was put in, filled with rubble. “That, together with removing 400,000 cubic yards of soil from the top end of the slip, seemed to stop the movement,” Mike says.
This Google Earth image shows the Broomhead reservoir. The slope that caused the problems is the forested area on the far side of the lake:-
The problems at this site are documented in some detail in a paper that was published in the Journal of the Institution of Civil Engineers (Beldelow 1944). This paper names the landslide the “Waldershelf Slip”. The article documents two phases of movement, with the first starting in 1924 during the construction of the overflow works (which can be seen in the image above). Beldelow (1944) includes the following cross section through the site – the slip can be seen on the right side of the diagram, with the inferred slip plane. Note the location in relation to the level of the top of the embankment:-
To stabilise the landslide, the engineers constructed a drain to extract water from the landslide and they removed about 300,000 m³ of material from the upper part of the slope to reduce the driving load. This work was completed in 1928, and Bendelow (1944) reports that the works appeared at first to have been successful.
However, in April 1930 further movement was observed on the slope, including cracks through the overflow weir. A very detailed investigation was undertaken of the causes of the instability, including trial pits dug to a depth of up to almost 30 metres! These found evidence of slip planes through the bedrock, indicating that the slope consisted of an ancient landslide that had been reactivated by the construction of the dam and associated structures.
To stabilise the slope, the engineers removed a further 125,000 m³ of material, installed drains within the landslide mass, and constructed drains beyond the margins of the landslide to stop water from entering the unstable part of the slope.
To the huge credit of the engineers, these measures proved to be entirely successful, and 75 years after the paper was published the reservoir remains in use.
Bendelow, L. 1944. Remedial works in connexion with the Waldershelf slip – Broomhead Reservoir. Journal of the Institution of Civil Engineers 22 (6), 95-106.
11 September 2019
Xe Pian Xe Namnoy: Land stability and dam failure on the Bolaven Plateau, Laos
A guest post by Richard Meehan and Douglas Hamilton
This progress report was prepared one year after the failure of the Xe Pian Xe Namnoy dam in Southern Laos and represents ongoing research into the causes of that failure by the authors. Early post failure studies were carried out by former dam designer Richard Meehan ([email protected]) with engineering geologist Douglas Hamilton, joining more recent efforts to consider implications for the future of the Xe Pian Xe Namnoy project and of dam building generally in similar upland areas of the Mekong basin.
The mountainous terrain of landlocked Laos creates a natural buffer between its modernizing and expanding neighbors, China, Vietnam, and Thailand. In recent times Laos has been recognized as a potential hydropower asset which could modernize the country and its remote upland populations. The high topographic relief and rainfall prevailing over much of the country are attractive for the development of high-head power drops but the monsoon-driven summer timing of rain requires seasonal reservoir storage near the edges of high plateaus. The capital-short Lao government has exploited these inviting conditions by opening the country to external hydropower investment and design/construction management by its neighbors. However such ambitious plans to turn the country into a profitable “battery of southeast Asia” have stumbled on problems arising from cases of hastily built and inadequately supervised construction in remote areas on what has often proven to be unstable ground.
A notable recent example is the July 2018 failure, on first filling, of a billion cubic meter reservoir, constructed for the Xe Pian Xe Namnoy project located on the Bolaven plateau in southern Laos. An initial review of this failure by the first author were presented in late 2018, and was followed six months later by a review by an independent expert panel drawn by the Lao government from the International Committee on Large Dams (ICOLD). Both reviews concur in finding that the failure was caused by a foundation failure beneath one of the project saddle dams.
On further review by us it appeared that the geomorphologic evolution of the high-altitude terrain that provides nominally favorable dam sites also features geologic defects typical of tropical areas that tend to make the ground and supporting structures unstable. Thus we find evidence of a form of generic geologic ground instability that may exist for similar sites of dam construction, existing and proposed, in this region.
The 2018 failure received international news coverage and commentary, such as an article in the Asia Times.
A general representation of the terrain of the Bolaven plateau presents a complex pattern of planation, uplift or erosional isolation, development of a trilinear joint pattern, weathering and erosion, followed by volcanic eruption and partial overlay of lava on the older sandstone:-
If effectively dammed, the network of narrow canyons on the older rocks on the southern part of the plateau provides some limited reservoir storage. More recent volcanism has partially filled some of the lower reaches of these canyons with lava flowing from surface vents flooding down from the north. The high secondary permeability of lava and possibly the residual laterite soils that form on it lead to sink drainages. This in turn results in groundwater flows toward the edges of the plateau, and where at points downslope these flows converge and rise, surfacing of groundwater flow creates a dendritic pattern of canyons cut into the easily weathered basalt lavas. At the edge of the plateau, the groundwater feeds notable seasonal waterfalls. At the upstream margins of the lava flows, remnants of weathered basalt form surficial canyon blockages that appear to be topographically favorable for dams but are deeply weathered to typical lateritic tropical soils. It now appears that this assemblage presents potentially unstable foundation conditions for dams.
The first author developed an interest in the failure as a result of having once in the 1960s participated in the design and construction of a problematical dam at a site with similar geological conditions but located on the uplifted southern rim of the Korat Plateau 500 km to the west. That dam was American designed based on experience with many projects in the western US but reflected lack of experience with tropical geology. The result was a near accident similar to the 2019 saddle dam failure.
With this interest and background, following the 2018 failure the first author made application for access to data on the Xe Pian Xe Namnoy project. But the Lao government and the Thai and Korean investors and engineers involved had quickly imposed a general blackout on all information on the project that continues even today in mid 2019, a year after the failure, even as the reservoir, with partial repairs, is now reportedly scheduled for refilling to recover financial goals of the $ 1 billion investment.
Initial simplified diagnosis of the failure was facilitated by a Google Maps terrain view of the site (15.008, 106.653) which shows a sink drainage above the dam. Details are shown in the author’s diagnosis above but can be replicated in a minute or two by the reader by entering those coordinates in a web browser and setting Google Maps to terrain view.
Access to the failure site has been prohibited but regional geologic mapping developed cooperatively by Lao and Japanese geologists is available. A clearer picture of the local physical geography emerged via scalable textured visualizations including the geological mapping and combined with SRTM and ALOS (approximately 2005) satellite data. Such a representation is shown here for the Bolaven Plateau, figure 2. The model was developed in QGIS and exported for 3-D visualization to Blender. The surface texture was processed by applying a high pass filter to the same terrain data and draping it as a brown-and-white image onto the 3-D manipulatable Blender model with exaggerated vertical scale. This preserves the texture of the local topography, making it relatively easy to discern between rough old sandstone, to the south, and smooth lava flows, to the north. Figure 3 shows a closer view of the same terrain model with project reservoir, dams, and mapped geology distinguishing Mesozoic sedimentary rocks from overlying lava flows.
With project data remaining secret additional important geologic and hydrologic data were needed and found in various Japanese studies undertaken initially in the early 2000 decade in connection with bauxite exploration and for several proposed dam sites, all of which are located on the margin of the lava flows and all questioned as to suitability, probably because of the low ground water and high seepage potential of the basalt. Among those sites so identified and proposed was a site for the main dam of the Xe Pian Xe Namnoy project; investigations demonstrated some problems including one basalt abutment with a low groundwater table and high seepage potential verified by borehole pressure tests. The final dam site was subsequently shifted upstream to its present location presumably to avoid these conditions. All of this information is available in online reports by the Japanese aid agency and served to make a possible diagnosis without specific detailed site investigation data at Xe Pian Xe Namnoy, assuming such data even exist.
A closer view of the portion of the reservoir including the main and three saddle downs is shown in Figure 3. The area mapped as basalt by Japanese and Lao geologists is indicated in by purple shading and is a good match with the geology pattern inferred from the terrain texture modeling described above. The upper reaches of the lava at about elevation 850 m form 10 km of the rim of the reservoir.
The inferred evolution of the terrain at the failed saddle dam site is shown in graphic sketches (Figure 4):-
With 3000 mm of annual summer rainfall it is evident that the underlying basalt was always able to convey a substantial subsurface flow beneath the saddle dam site, opportunistically located on a remnant basalt/laterite ridge lacking development of surface drainage. Here as at the other dams at Xe Pian Xe Namnoy and at yet more dam sites proposed for the Bolaven Plateau, the flow of groundwater in the basalt passes toward the edge of the plateau (in this case to the southwest), eventually emerging to the surface and creating dendritic pattern of incised canyons that is quite different from the trilinear joint pattern seen in the Mesozoic rocks.
The Japanese engineering geologists in their investigations have identified lava cooling joints as principal seepage paths, but other conditions promoting high and concentrated subsurface flow and erosion including those that might be associated with hot lava flowing in wet jungle canyons — volcanic tubes, brecciated zones, perhaps maar-like features, as well as uncompressed or open stress relief joints in the underlying sandstone can be readily proposed. Runoff deficits from sink areas including at the saddle dam site are quite large, consistent with experience elsewhere with losing streams in lava areas. Borehole water pressure tests taken at various of the rejected other sites on the plateau indicated high flows often exceeding the pumping capacity of the equipment.
A telling pair of photos of the dam crest a few hours before failure is shown in Figures 5a and 5b:-
Both downward and lateral deformation of the dam are evident. The first author’s sketch of one of the hypotheses of the generation of this failure pattern is shown on figure 6. Note that the dam embankment does not include any drains or zoning that would protect against seepage failure of the embankment or relieve foundation water pressures. Alternative or complementary foundation seepage paths including eroded tunnels and cavities developing probably at the top of the basalt or between the laterite and basalt, or the basalt and the upper sedimentary rocks can also be imagined. Unprecedented uplift pressures beneath the cap of laterite developed. A large component of ground slippage toward the drone camera, i.e. away from the abutment, is suggested in figure 1. The general concept of a seepage induced foundation failure was initially advanced by the writer in October 2018 three months after the failure. The official government expert panel reportedly came to the same findings, but this report has never been released, reportedly out of deference to the Korean government.
A more ominous scenario for the 2018 failure involves the partial mobilization of the ridge area at and downstream of the left abutment of the dam along with a simultaneous failure at the saddle dam. This possibility is suggested both by the canyon-facing scarps in Figure 1 and also by the presence of a separate landslide on the left side of the canyon on ground not directly impacted by the saddle dam (Figures 7 and 8):-
Reportedly all of the saddle dams are to be replaced by new roller-compacted concrete dams. But adequate stability of the other two saddle dam foundations would surely need to be verified to allow this, considering the precedent of the famous 1928 St. Francis Dam failure. That too was taken to be superior in strength by virtue of its being made from concrete. But the St. Francis dam was fatally sited on a weak and weathered foundation (Figure 10). This case history renders the reported plan, if true, for construction of “strong” concrete dams on weak lava-derived foundations highly questionable.
If, as mapped by the Japanese mining geologists, and suggested as well by the textured 3D models, the 10 km southwestern rim of the reservoir is underlain by deep laterite soil, leaky basalt, and weathered sedimentary rocks, the suitability of concrete dams must be considered. Moreover, a pattern of unstable ground could extend widely beneath the entire ridge between the reservoir and the canyon below the failed saddle dam. Generalized landslide slope failures along this ridge can be suspected from photographic evidence as shown in Figure 9. Immediate failure of the ridge might not occur because time would be required following filling of the reservoir to raise the water table to critical levels. A major failure of the ridge could be a greater catastrophe than the 2018 failure.
The severe social impact of the project and its failure last year have received censure from the UN and other parties. As of mid 2019 the viability and future of the Xe Pian Xe Namnoy project remains uncertain.
Richard Meehan and Douglas Hamilton. All views expressed are those of the authors.