The team combined field observations, satellite imagery, and two-dimensional hydrodynamic simulation to reconstruct the 12 July 2018 outburst from the Kunj-Peshghor lake, which killed four people, destroyed more than forty houses, and swept away irrigation canals and infrastructure in Peshghor village. The flood struck at midnight. Shepherds on the mountainside, seeing the surge coming, fired gunshots to warn villagers downstream, but many residents were asleep and never received the message. The flood wave from the Peshghor tributary carried so much sediment into the main Panjshir river that it temporarily dammed the larger channel, creating a lake that backed up 1.7 kilometers upstream from the confluence.

What makes the 2018 event scientifically remarkable is how quickly the lake itself formed. Satellite records show the Kunj-Peshghor lake appeared around 2 June 2018 on the surface of a debris-covered glacier and expanded rapidly over just 42 days before draining catastrophically. The 2021 Bam-Tanab event followed an almost identical script: a supraglacial pond formed and grew over roughly 33 days before breaching on 4 July, damaging twelve houses, eight irrigation canals, four micro-hydropower plants, and a rural water supply network. Both lakes belong to a class the authors call debris-ice glacial lakes, unstable water bodies dammed by mixtures of ice and rock debris that can vanish overnight.

The triggers were intensely local. In 2018, sustained high air temperatures, very low humidity, and an absence of rainfall drove rapid snow and ice melt that filled the lake, before water eroded a subglacial tunnel beneath the glacier and drained it from below. In 2021, extreme heat combined with a significant rainfall event five days before the outburst, which swelled inflows into both the pond and the main lake until an englacial tunnel roughly five by seven meters widened on the lake’s north side and released about 25,000 cubic meters of the stored 33,400 cubic meters. Field reports noted the collapse of around five meters of surrounding ice as the conduit enlarged.

To quantify what such a flood does on its way downhill, the researchers coupled the BASEbreach dam-breach module with the open-source BASEMENT two-dimensional hydrodynamic model, solving the shallow water equations on an unstructured triangular mesh built from a 10-meter digital elevation model, resampled to 14 meters for computation. Building footprints digitized from pre-event Planet imagery were raised by five meters in the terrain model so the mesh represented houses as flow obstacles. The model domain was divided into seven land-cover classes, from riverbed and floodplain to roads, crops, and trees, each assigned its own Strickler friction coefficient ranging from 10 for trees to 60 for roads.

Calibration against the flood extent mapped from 3-meter Planet imagery one day after the disaster identified an optimal roughness combination with a precision of 0.95 and an F1-score of 0.990. Because no gauging stations exist along the valley, validation relied on flood marks left on mosque walls, containers, and a car, which indicated depths of 1.6 to 2.0 meters in the worst-hit areas; simulated depths differed from these observations by only about 0.2 meters on average. The empirical breach analysis yielded a peak discharge of roughly 507 cubic meters per second, delivered as an asymmetric hydrograph rising to its peak within about 2,000 seconds and receding over nearly three hours, with a total released volume of about 880,000 cubic meters.

The simulations reveal a flood that is anything but uniform. Near the source, depths were modest at around two meters, but as the flow funneled through constricted valley sections, velocities peaked at 7.5 meters per second at roughly 6.5 and 11 kilometers downstream, and depths exceeded 10 meters in the mid-reach between 8.5 and 9 kilometers. In the lower valley near Peshghor village, velocities of about six meters per second persisted with depths of 2.5 to 6 meters. Sensitivity tests showed that doubling or halving channel friction shifted peak depths between 9.2 and 13.1 meters and peak velocities between 8.6 and 9.5 meters per second, while total inundated area changed by only about six percent, suggesting flood extent alone is a poor calibration target and that depth and velocity distributions matter more.

Downstream hydrological records add a second, often overlooked act to the disaster. At the Tangi-Gulbahar monitoring station, turbidity spiked above 800 NTU immediately after the 2018 outburst and declined only over several days, while daily discharge jumped from 75 to over 160 cubic meters per second in 2018 and above 170 in 2021. Both stations recorded two distinct water-level peaks: the first from the original flood surge, the second when debris temporarily blocked the Panjshir river and then released. The river channel itself widened from 30 to 50 meters before the flood to nearly 250 meters in places, with more than 200 meters of channel choked by flood debris. These secondary surges prolong flooding, degrade water quality, and complicate emergency response in ways that single-peak hazard models miss.

Overlaying simulated inundation on land-use data showed that roughly 46 percent of the flooded area in Peshghor village falls within high to very high hazard zones, including homes, the main mosque and madrasa, the district road, market areas, and agricultural land. Very high hazard zones cover about 18 percent of the inundation, concentrated along the main channel, with high hazard zones adding another 28 percent. The simulated extent closely matched the documented 2018 damage: about 80 houses affected, 40 destroyed, roughly 20 hectares of cultivated land lost, 400 meters of irrigation canal, two bridges, 1.5 kilometers of road, and one micro-hydropower facility. Villagers rebuilt in the same flood zone after 2018, only to be hit again in 2021, a pattern that underscores how exposure persists when hazard maps do not exist.

The broader implication is uncomfortable for the field. Conventional GLOF hazard assessments rely on remote-sensing inventories of long-lived, moraine- or bedrock-dammed lakes, rating their susceptibility from dam characteristics and surrounding conditions. But lakes that form and fail within weeks, draining through hidden englacial tunnels that are nearly invisible from orbit before collapse, escape such screening entirely. Between 1990 and 2020 the number of glacial lakes in Afghanistan’s Hindu Kush rose 32 percent to around 2,600, with 17 percent newly formed, and recent assessments flag more than 36 lakes as highly susceptible. With roughly one million people across High Mountain Asia living within 10 kilometers downstream of dangerous lakes, the authors argue for high-frequency monitoring of debris-covered glaciers, automated lake-level loggers, seismic and acoustic sensors near breach zones, telemetry-linked warning systems that could extend traditional gunshot alerts, and hazard zoning that keeps new construction off the terraces the next flood will claim.

Subject of Research: Glacial lake outburst flood hazard, dam-breach dynamics, and downstream impacts in the Hindu Kush of Afghanistan

Article Title: Glacial lake outburst flood hazard and downstream impacts: lessons from Afghanistan examples

Article References: Azizi, F., Fatihi, A., Antoniazza, G., Bertolazzi, M., Ahmed, R., & Lane, S. N. (2026). Glacial lake outburst flood hazard and downstream impacts: lessons from Afghanistan examples. Natural Hazards, 122(21), Article 665. https://doi.org/10.1007/s11069-026-08449-3

Image Credits: AI Generated

DOI: 10.1007/s11069-026-08449-3

Keywords: glacial lake outburst flood, GLOF, Afghanistan, Hindu Kush, Panjshir, hydrodynamic modeling, BASEMENT, dam breach, debris-covered glacier, climate change, early warning systems, flood hazard mapping

The team, led by Ankita Ghoke, Priyom Roy and colleagues, combined three complementary lines of evidence to track the slope’s evolution. Eighteen cloud-free optical images from Indian Resourcesat-2 and 2A satellites, the European Sentinel-2 mission and the high-resolution Cartosat-3 sensor allowed them to map morphological changes between 2013 and 2025. They then applied multi-temporal interferometric synthetic aperture radar, or MT-InSAR, using the Small Baseline Subset technique on Sentinel-1 radar data to measure millimeter-scale ground deformation over four distinct time windows. Finally, field campaigns after the 2024 reactivation documented the structural and geological weaknesses that make the slope so prone to failure. Together, these datasets reveal a landslide that is not a single event but a slow-motion, retrogressive collapse that keeps climbing uphill and swallowing new ground.

The optical record tells a striking story of escalation. In March 2013, the unstable zone was a modest, elongated feature roughly 500 to 600 meters long. A major reactivation in June 2013, coinciding with the disastrous Kedarnath cloudburst that devastated much of Uttarakhand, stretched the slide to about 965 meters and carved the first prominent head scarp while blocking the Shambhu River. Between 2014 and 2016 the failure continued to retreat upslope, damming and re-damming the river. Around October 2018, an entirely new landslide initiated to the southwest, and by late 2019 its scar had merged with the main slide mass, enlarging the failure zone still further. By 2024, four separate head scarps were visible on the slope, evidence of multi-scarp retrogression still underway. The landslide now covers approximately 0.95 square kilometers, and imagery shows that several settlements in its path were destroyed as it grew.

The radar measurements add a crucial dimension: they show the slope moving before it visibly fails. In the first analysis phase, from October 2014 to October 2016, monitoring points on the flanking slopes recorded average line-of-sight velocities of roughly 40 millimeters per year, with deformation accelerating during the 2015 and 2016 monsoons before complete failure and scarp formation. In the second phase, a point tracking the nascent secondary landslide crept at 13.6 millimeters per year before a rapid acceleration culminating in the October 2018 collapse. Later phases recorded velocities between about 5 and 12 millimeters per year, each acceleration episode closely following periods of intense rainfall. The pattern is classic transient landslide behavior: stability, then primary creep, then accelerating motion, then failure, with monsoon precipitation acting as the trigger on a slope already predisposed to collapse by its geology.

That geology is unfavourable in almost every respect. The slope lies near the Main Central Thrust in the seismically active Lesser Himalaya, underlain by highly sheared talc schist, carbonaceous phyllite, quartzite and slate of the Lameri Formation, all draped in thick, loose colluvium. Fieldwork revealed foliation planes dipping at 45 to 50 degrees, parallel to the slope face, creating ready-made planar failure surfaces. The team also mapped transverse cracks, active groundwater seepage channels exuding along those cracks, and linear discontinuities that appear to be faults or joints. Particularly worrying is the village of Kuwari itself, which sits on a previously stabilized portion of the landslide body that field evidence now shows is bulging and deforming anew. A road constructed across the secondary landslide zone in early 2022 has further destabilized that section, a reminder that human activity can amplify an already fragile situation.

Meanwhile, the lake behind the debris dam keeps growing. The first impoundment formed in June 2013 and drained in 2015; a second formed in November 2016, partially drained in October 2017 and re-formed in 2018. From November 2021 onward the water body expanded relentlessly: roughly 885 meters long by April 2022, about 900 meters by October 2022, 1,050 meters by October 2023, and approximately 1,130 meters long and 100 meters wide by October 2024, the largest extent on record. Each enlargement reflects fresh debris choking the valley and raises the volume of water poised above downstream communities. History offers little comfort: previous research indicates that 80 to 90 percent of landslide-dammed lakes fail within their first year, and the Kuwari dam has already been partially breached by the river cutting a narrow channel through the debris.

To assess how stable that dam actually is, the researchers applied two geomorphic indices. The Morphological Obstruction Index, calculated at 3.82, places the dam in the uncertain evolution domain, while the Hydro-morphological Dam Stability Index, computed from the dam volume, upstream catchment area of about 2.46 square kilometers and channel gradient, falls squarely within the instability domain. The dam itself is a type III barrier in the classification of Costa and Schuster, meaning debris has moved both upstream and downstream of the valley and completely obstructs it. With an estimated height of around 40 meters and a surface area near the crest of roughly 10,000 square meters, the structure is an unconsolidated pile of landslide debris, not engineered material, and it is already being eroded by the river flowing through it.

The most dramatic part of the study is the flood simulation. Using HEC-RAS version 6.7, the team modeled a complete overtopping breach of the dam, drawing breach geometry from an empirical relationship derived from historical dam failures. Assuming the full 3.7 million cubic meter reservoir and a 17-meter breach height, the model predicts an average breach width of 48.3 meters forming in just 0.64 hours, less than forty minutes. The resulting flood would race down the narrow valley at a peak velocity of 15 meters per second near the breach, with overbank flooding extending about six kilometers downstream to Choting village, where water could rise up to 10 meters above normal levels and inundate agricultural terraces. Beyond Choting, the flow would remain channelized but still dangerous, reaching the villages of Melkhet, Deosari and Dewal at average velocities of 6 to 8 meters per second, fast enough to threaten anything along the riverbanks.

The authors are careful to note that the simulation is a predictive, end-member scenario that cannot yet be validated, and that its results carry uncertainties in breach width, formation time, erosion rates and dam material properties. But the broader message is hard to escape. The Kuwari landslide is an active, compound hazard in one of the world’s most landslide-prone countries, which a global analysis by Froude and Petley ranked highest in landslide fatalities. The study’s authors call for continuous monitoring, an early warning system for downstream communities, slope stabilization measures, and strict controls on construction across already scarped slopes. As climate change intensifies monsoon extremes across the Himalaya and infrastructure development pushes further into fragile terrain, the integrated approach demonstrated here, pairing satellite radar with optical imagery, field geology and hydrodynamic modeling, may become an essential template for spotting the next Kuwari before its lake bursts.

Subject of Research: Reactivation dynamics of the Kuwari landslide in Uttarakhand, India, and assessment of landslide lake outburst flood hazard using remote sensing and hydrodynamic modelling.

Article Title: Reactivation dynamics of the Kuwari landslide (Uttarakhand, India) and assessment of LLOF hazard using optical imagery, MT-InSAR and hydrodynamic modelling

Article References: Ghoke, A., Gaur, A., Roy, P., Jalan, P., Martha, T. R., & Das, I. C. (2026). Reactivation dynamics of the Kuwari landslide (Uttarakhand, India) and assessment of LLOF hazard using optical imagery, MT-InSAR and hydrodynamic modelling. Discover Geoscience, 4(1), Article 345. https://doi.org/10.1007/s44288-026-00717-w

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00717-w

Keywords: Kuwari landslide, Uttarakhand, Himalaya, MT-InSAR, landslide-dammed lake, LLOF, HEC-RAS, Sentinel-1, dam breach, natural hazards, remote sensing, monsoon rainfall

The instrument, which the researchers call a computer vision-based measurement monitoring (CVMM) system, works on a principle reminiscent of a surveyor’s theodolite. A 20-centimeter chessboard target is fixed to the point of interest on the dam crest, while the camera sits at a stable vantage point on the opposite riverbank. OpenCV algorithms running in Python detect the corners of the chessboard in each frame, calibrate for lens distortion, and convert pixel shifts into real-world displacement using the mathematics of planar homography and camera projection matrices. Because the true size of the chessboard is known, the system can even estimate the distance between the camera and the target from how large the board appears in the image, recovering motion along the depth axis as well as horizontally and vertically.

Accuracy was established through careful laboratory characterization. On a three-axis manual stage, the static standard deviation of the measurements was just 0.058, 0.042, and 0.067 pixels in the X, Y, and Z axes respectively, across more than 2,500 data points per axis. When the chessboard was moved by exactly one centimeter, the system reproduced the motion with an accuracy of 0.01 centimeters, where a single pixel corresponded to roughly 0.16 centimeters at the test distance. Long-range trials at 15, 20, 35, and 50 meters showed average standard deviations of 0.01 to 0.02 centimeters up to 35 meters, degrading to 0.06 centimeters beyond 50 meters, which led the team to recommend installation distances under 35 meters for their industrial lens with 10x optical zoom.

Crucially, the system does more than track displacement. Running at 100 frames per second—nearly the maximum 120 frames per second supported by the camera module—the CVMM captures vibrational motion that conventional time-lapse photography would average away. Displacement time series can be differentiated to yield velocity, and then decomposed spectrally. The researchers applied the Fast Fourier Transform to isolate dominant frequencies, the Short-Time Fourier Transform to see how those frequencies evolve in time, and finally the Hilbert-Huang Transform, a technique designed for nonlinear, non-stationary signals that breaks the record down into intrinsic mode functions and computes instantaneous frequency and energy for each one.

To validate the vibration measurements, the team mounted a chessboard alongside a commercial Raspberry Shake 4D seismograph on a QUANSER shake table and drove the table through a series of controlled oscillations at constant 1-centimeter amplitude but varying accelerations, from 0.8 to 200 gal. The image acquisition frame rate achieved 98.76 percent of its 100-frames-per-second target, and image recognition succeeded 99.70 percent of the time. After a median filter smoothed out spike noise, spectral analysis showed that the primary frequencies detected by the CVMM aligned closely both with the shake table’s input settings and with the independent seismograph record. The maximum amplitude error was 0.13 centimeters, with an average of 0.09 centimeters, confirming that a camera tracking a checkerboard can genuinely function as a non-contact vibrometer across a range from 0.1 to 50 Hz.

The decisive test came on the Lantao river in Nantou County, Taiwan, where the researchers built an artificial earth dam 27.8 meters wide and 3.0 meters high from a mix of boulders, gravel, and sand typical of landslide-formed debris dams. The chessboard was planted on the dam crest near a pre-cut breaching channel, and the CVMM instrument was set up 35.32 meters away on the right riverbank, a distance verified with a theodolite. Water was then fed into the reservoir until it overtopped the crest and carved the dam apart. Over 1,249 seconds of monitoring, the target slid 14.64 centimeters horizontally toward the developing breach and settled 26.77 centimeters vertically before it finally collapsed with the failing dam.

The time-frequency analysis of this field data revealed a striking pattern. As the breach developed between 15:16 and 15:20, vibration energy surged across three distinct frequency bands: a high-frequency band from 20 to 50 Hz, a medium band from 10 to 20 Hz, and a low band below 10 Hz. Following established findings that high-frequency ground vibrations arise from small particles like sand and gravel colliding while low frequencies reflect the movement of larger rocks, the team concluded that the dominant energy came from the washing out and collision of finer sediments as overflow eroded the crest. The vertical axis carried more energy than the horizontal—maximum instantaneous energies of 2,500, 350, and 50 square centimeters per second squared in the high, medium, and low bands of the Y axis, compared with 1,800, 200, and 30 on the X axis—consistent with settlement being the more violent process.

Most importantly for disaster response, the Hilbert spectra showed a sharp rise in instantaneous energy beginning around 15:16, roughly two minutes before the overflow at 15:18 and the full breach at 15:20. Because the first intrinsic mode function carried over 84 percent of the total energy, the researchers proposed an early-warning threshold of 1,000 square centimeters per second squared in that high-frequency band, a level that was exceeded more than 36 times during the initial breach phase. Combined with a simpler displacement indicator—settlement exceeding 5 centimeters at the moment of overtopping—these thresholds could give downstream communities precious minutes of advance notice. LiDAR scans taken before and after the test independently confirmed the elevation changes the camera system had tracked in real time.

The economics are equally compelling. The entire CVMM instrument costs roughly 1,000 to 2,000 US dollars depending on configuration, about one-fifth the price of conventional monitoring packages such as total stations, GNSS receivers, or LiDAR units. If the chessboard target is destroyed in a failure event, replacing it is trivial. The system also delivers three data streams simultaneously—imagery, displacement, and vibration—where each traditional device provides only one, and it requires no physical contact with a structure that may be on the verge of collapse, addressing a fundamental weakness of contact sensors like the Raspberry Shake, which cannot safely be deployed on a failing slope.

The authors are candid about the limitations. The energy threshold was derived from a single field test on one soil type and will need validation across different dam materials. Computer vision remains vulnerable to heavy rain, poor night lighting, atmospheric turbulence, and camera shake, although previous work by the same group using a weatherproof enclosure and a solar-powered LED chessboard achieved 24-hour monitoring with errors near 0.1 centimeter at 50 meters. No Raspberry Shake was deployed on the dam crest itself for direct comparison during the breach, and coupling the vibrational signatures to pore pressure changes and internal failure mechanisms will require additional instruments such as piezometers. Even so, the demonstration stands as a proof of concept that a modest camera, a printed checkerboard, and clever signal processing can transform dam safety monitoring—turning an ordinary machine vision pipeline into an early-warning sentinel for one of the deadliest failure modes in mountainous terrain.

Subject of Research: Computer vision-based monitoring of dam breach displacement and vibration for early warning

Article Title: Computer vision-based system of vibration monitoring for early-warning signals in dam breach

Article References: Chen, I.-H., Chen, S.-C., & Yang, R.-J. (2026). Computer vision-based system of vibration monitoring for early-warning signals in dam breach. Environmental Earth Sciences, 85(15), Article 392. https://doi.org/10.1007/s12665-026-13117-7

Image Credits: AI Generated

DOI: 10.1007/s12665-026-13117-7

Keywords: computer vision, dam breach, vibration monitoring, displacement monitoring, Hilbert-Huang Transform, OpenCV, Raspberry Pi, early warning, time-frequency analysis, earth dam, landslide dam, structural monitoring