Construction site of the Nam Tha 1 dam in Bokeo Province, Laos, July 2017 (Source: CNN)
Bokeo Province, Laos
The geography of the south-east Asian country of Laos is dominated by rugged mountains, giving Laos significant capacity for hydroelectric power generation. Laos has been dubbed the “Battery of Asia” and is aiming to have the capacity to create 30, 000 MW of hydroelectric energy by 2040.1
The Xe-Pian Xe-Namnoy power project in south eastern Laos is a $1.2B USD project that started construction in 2013. The project has two primary dams and several subsidiary structures known as saddle dams, which assist in containing the reservoir.2 On July 23, 2018, Saddle Dam D collapsed, unleashing a wall of water on the villages downstream. In total, an estimated 5bn cubic litres of water flooded the region, removing entire villages in its path.3 The flash flood resulted in numerous deaths and forced several thousands of victims to lose their homes. Figure 1 shows the depletion of the reservoir and the creation of new water channels flowing from the reservoir, as illustrated from Sentinel-1 SAR amplitude images taken before and after the collapse.
1 “Is Laos facing a dam disaster? – CNN – CNN.com.” 14 Dec. 2018, https://www.cnn.com/2018/12/14/asia/laos-hydropower-dams/index.html. Accessed 22 May. 2020.
2 “How the Laos dam disaster unfolded.” 26 Jul. 2018, http://fingfx.thomsonreuters.com/gfx/rngs/LAOS-ACCIDENT/010071FR2TL/index.html. Accessed 22 May. 2020.
3 “How the Laos dam disaster unfolded.” July 26, 2018, http://fingfx.thomsonreuters.com/gfx/rngs/LAOS-ACCIDENT/010071FR2TL/index.html. Accessed May 22, 2020.
Figure 1: SAR amplitude images acquired by the Sentinel-1 satellite on July 17, 2018 (left) and July 29, 2018 (right), corresponding to dates before and after the collapse of Saddle Dam D. Water appears as dark patches in the images, with the reservoir shrinking drastically between the two images.
- Was there any precursor displacement on the dam detectable at significant levels in archival InSAR databases?
- Could an operational monitoring project, operating without the benefit of hindsight, have produced actionable intelligence for the dam operators warning of impending collapse?
- The surrounding jungle outside the dam is highly incoherent at C-band.
Solution & Results
3vGeomatics obtained 37 C-band Sentinel-1 SAR images taken between May 11, 2017, and July 17, 2018. These images were collected with descending passes, when the Sentinel-1 satellite was travelling north to south and looking down and to the west. Such a descending pass provides the most sensitivity to downslope motion along the west facing Saddle Dam D.
3vGeomatics has developed a processing chain tailored to monitoring coherent dam faces in forested areas, where the coherence is low in most areas outside of the dam. This tailored dam monitoring chain was used to isolate and process high quality, coherent points in and around Saddle Dam D.
The InSAR analysis shows that immediately after construction, Saddle Dam D began to settle with the period of settlement continuing until early November 2017. From November 2017 until May 2018, the dam showed little to no displacement. Starting in early May 2018, roughly three months before the collapse of the dam, the dam began to undergo accelerating displacement. By July 17, 2018, sections of the dam had undergone over 3 cm of displacement and were subsiding at rates greater than 20 cm / year. Figure 2, right, shows the displacement measured from November 2017 until July 17, 2018, just prior to the dam’s failure.
3vGeomatics provides signal-to-noise contours on its products, which highlight areas with statistically significant displacement. On the Laos dam, displacement is detected at a 4-sigma level (four times the noise floor, 99.99% confidence interval) in one location, and at a 3-sigma level (99.7% confidence interval) in three additional locations along the dam. These areas are indicated by the turquoise and white contours, respectively, in Figure 2.
3vGeomatics has an automated tool to detect significant changes in the displacement rate, i.e. rapid acceleration or deceleration. Contours showing areas with significant acceleration are included in standard products. Over Saddle Dam D, significant acceleration was detected at a 2-sigma level, shown as the black contour in Figure 2.
Figure 3 shows the time series derived for this section of the dam, with the accelerating subsidence of the dam clearly evident. A visual inspection of the final 12-day interferogram also shows clear displacement signals on the dam, which would have been flagged by 3vGeomatics trained operators.
- Actionable alerts regarding the presence of displacement
- Actionable alerts regarding the acceleration of displacement
- Rates and accelerations quantitatively classified with significance thresholds to clearly delineate real signal from noise
- Possible early warning allowing for mitigation strategies to be used to prevent catastrophic failures
This failure highlights a greater need for risk mitigation procedures of hydroelectric dams, not only in Laos, but globally. Beyond the requirement for high factors of safety built into a dam’s design, coupled with quality control procedures and visual inspections, there is a demand for measuring the performance of a dam through instrumentation and monitoring. Within an effective dam instrumentation and monitoring program, InSAR would have the ability to monitor subtle movements of all dams, reservoir slopes, and subsidiary dams on a 2 to 12 day frequency. This would provide the necessary data for dam operators to make potential design / construction adjustments.
3vGeomatics has developed scalable processing chains for monitoring dams globally and would likely have been able to provide early warnings that Saddle Dam D was undergoing accelerating subsidence if 3vG were to have been monitoring the dam. For further information regarding 3vGeomatics’ dam monitoring capabilities, please feel free to contact us.
Figure 2: Displacement rates derived over Saddle Dam D from November 2017 until July 17, 2018, overlaid over Sentinel-2 optical imagery of the dam. The turquoise and white contours denote areas with 4-sigma and 3-sigma detections of displacement, respectively. The black polygon shows an area where significant acceleration was measured.
Figure 3: Time series of subsidence in the area of Saddle Dam D with significant acceleration detected. The black lines show time series for individual points on the dam while the red line shows the average of all points.