The new study, published in the journal Natural Hazards by Yiwei Liu, Aiguo Xing, and colleagues at Shanghai Jiao Tong University, the China Institute of Geo-Environment Monitoring, and Hunan University, presents a joint analytical framework that fuses two very different data streams: ordinary video recordings and seismic observations. The work is a case-oriented, exploratory attempt to integrate visual and seismic information for rock avalanche analysis, and it demonstrates its approach on a representative large rock avalanche event in Southwest China. The goal is deceptively simple to state but technically demanding to achieve: to obtain velocity distributions and motion descriptions at any moment during an avalanche, something neither data source can deliver alone.

The core of the visual analysis is a technique called optical flow, a cornerstone of computer vision with a history stretching back to foundational work in the early 1980s. Optical flow algorithms estimate the apparent motion of pixels or features between successive frames of a video, producing a dense field of motion vectors that describes how every part of the image is shifting. In the context of a rock avalanche, those vectors translate directly into kinematic information: which parts of the flowing mass are moving fastest, where the front of the avalanche is advancing, and how the body of the flow stretches, compresses, and deforms as it descends. The researchers implemented their pipeline in Python, building on widely used open-source libraries including OpenCV, NumPy, and pandas, and they have released the code and data openly through GitHub and the Zenodo repository.

Video alone, however, has a fundamental weakness. A camera records only what is in its field of view, and the motion it captures is projected onto a two-dimensional image plane. Without independent confirmation, it can be difficult to know precisely when events in the video occurred in absolute time, or whether apparent motion reflects real ground movement or artifacts such as camera shake, changing illumination, or dust obscuring the scene. This is where the seismic half of the framework becomes essential. Large rock avalanches generate ground vibrations that are recorded by seismic instruments at considerable distances, and seismologists have developed increasingly sophisticated methods to invert those signals for landslide characteristics such as location, timing, momentum, and frictional behavior.

In the joint framework, seismic signal analysis serves as an independent reference for temporal alignment and validation. By matching prominent features in the seismic record, such as the abrupt onset of high-frequency energy when the mass impacts the valley floor, with corresponding moments in the video, the researchers can anchor the visual timeline to a precise, instrumentally measured clock. The seismic data then act as a cross-check: if the optical-flow analysis suggests a surge in motion at a particular moment, the seismic record should show a corresponding change in energy release. Agreement between the two independent streams increases confidence in the reconstructed dynamics; disagreement flags moments that deserve closer scrutiny.

The framework is organized into a four-stage architecture that integrates data acquisition, preprocessing, joint analysis, and result visualization. In the acquisition stage, video footage is gathered from monitoring systems and public sources, alongside seismic records covering the event window. Preprocessing addresses the practical realities of real-world imagery: stabilizing frames, correcting for lighting variation, and preparing the seismic traces through filtering and decomposition techniques. The joint analysis stage runs the optical-flow motion extraction and the seismic interpretation in parallel, aligning them in time. Finally, the visualization stage renders the results as velocity distributions and motion descriptions that can be examined at any moment during the avalanche, turning raw footage and seismograms into quantitative, interpretable pictures of the disaster in progress.

Applied to a large rock avalanche in Southwest China, the framework demonstrated its practicality. The results showed that the combined approach provides complementary insights to those obtained from seismic signal analysis alone. Seismic inversion can estimate bulk properties of a landslide, but it inherently averages the behavior of the entire moving mass. Video-based optical flow, by contrast, resolves spatial patterns: it can reveal that the front of the flow accelerates while the trailing mass decelerates, or that deformation concentrates along particular zones within the avalanche body. Merging the two perspectives yields a richer and more defensible reconstruction than either could offer independently.

The significance of this capability extends well beyond academic curiosity. Rock avalanches routinely cause severe casualties and economic losses, and improving the characterization of event-scale dynamic behavior and precursory deformation is considered essential for effective disaster mitigation and early warning. Landslides are among the deadliest geological hazards worldwide, and China, with its steep terrain, active tectonics, and dense population in mountainous regions, has experienced repeated catastrophic events, including the long-runout Pusa rock avalanche of 2017 in Guizhou Province and the massive Chamoli rock and ice avalanche in the Indian Himalaya in 2021, which killed hundreds. Understanding exactly how these masses accelerate, fragment, and spread is critical for designing warning systems, mapping hazard zones, and validating the numerical models engineers use to predict runout.

What makes the new framework particularly timely is the explosion of available video. Monitoring cameras now watch many hazardous slopes continuously, and when disasters strike, surveillance cameras, smartphones, and public platforms often capture the event from multiple angles. Until recently, this footage was mostly used qualitatively, as dramatic evidence of what happened. The authors argue that video-based analysis has become a valuable complementary data source for investigating landslide dynamics, and their framework provides a systematic, reproducible way to convert pixels into physics. Optical-flow methods have already proven useful in related domains, from measuring glacier surface velocities in satellite imagery to tracking slow slope creep in high-alpine settings, and this study extends that lineage to the most violent end of the landslide spectrum.

The researchers are careful to frame the work as a case-oriented and exploratory attempt rather than a finished operational system. Challenges remain, including the sensitivity of optical flow to dust, poor lighting, and low frame rates, and the difficulty of converting image-plane motion into true three-dimensional ground velocities without accurate camera calibration and terrain models. Nevertheless, the study highlights the potential of video-based computer vision techniques to enhance the interpretation of landslide dynamics and to support future monitoring and early-warning research. If the approach matures, the aftermath of the next catastrophic slope failure may be documented not just by trembling seismometers and stunned witnesses, but by algorithms that watched every frame, measured every surge, and helped explain one of nature’s most violent phenomena in unprecedented detail.

Subject of Research: Integration of computer vision and seismic signal analysis to characterize the dynamics of large rock avalanches

Article Title: Characterizing dynamic motion and deformation during large rock avalanches using a joint framework integrating computer vision and seismic signal analysis

Article References: Liu, Y., Xing, A., Wang, W., Wang, Q., Zhuang, Y., Bilal, M., & Zhu, K. (2026). Characterizing dynamic motion and deformation during large rock avalanches using a joint framework integrating computer vision and seismic signal analysis. Natural Hazards, 122(18), Article 628. https://doi.org/10.1007/s11069-026-08349-6

Image Credits: AI Generated

DOI: 10.1007/s11069-026-08349-6

Keywords: rock avalanche, optical flow, computer vision, seismic signal analysis, landslide dynamics, deformation characterization, early warning, disaster mitigation, video-based motion analysis, Natural Hazards, China, temporal alignment

The two landslides struck within minutes of each other, at approximately 3:50 and 3:53 UTC, from a source area with an initial estimated volume of about 280,000 cubic meters. As the sliding masses entrained surficial material along their paths, they left a debris deposit stretching roughly 1,162 meters in total length. The site lies in the geomorphic transition zone between the Sichuan Basin and the Yunnan–Guizhou Plateau, an area of stepped topography and deeply incised valleys where sharp elevation drops concentrate surface runoff and dissect the terrain. The local stratigraphy consists of Quaternary slope deposits underlain by the Lower Triassic Feixianguan Formation and the Upper Permian Xuanwei Formation, with a rock sequence exhibiting a reverse dip and dominated by interbedded sandstone and weaker mudstone.

To capture the event, the researchers turned to three nearby broadband seismic stations: JLI, YAJ, and HWS. Landslide-induced signals were identifiable at all three, but only JLI, the closest station, delivered clear, complete three-component waveforms with a high signal-to-noise ratio suitable for quantitative inversion. Using the classical short-term average to long-term average energy-ratio method, with a 0.5-second STA window, a 5-second LTA window, and a trigger threshold of 10, the team identified the first event’s detectable signal from 03:51:01 to 03:52:02 UTC, lasting about 61 seconds, and the second event’s from 03:53:48 to 03:54:29 UTC, lasting about 41 seconds. Because no seismic-wave travel-time correction was applied, these durations approximate rather than exactly date the physical movement, but they provide crucial temporal constraints on a process that would otherwise be invisible.

Time-frequency analysis using the Hilbert–Huang Transform revealed that the seismic energy of both landslides was concentrated below 3 hertz, a hallmark of large, coherent mass movements. The first landslide’s signals showed a striking bimodal pulse pattern, with distinct pulses at roughly 19 and 41 seconds after onset. Matching these pulses against field-mapped terrain features, the team interpreted them as impacts at two points along the runout path: one where the slope gradient dropped abruptly and the sliding mass slammed into the ground surface, producing a strong vertical vibration, and another where the terrain azimuth changed sharply, forcing a violent directional adjustment. The second landslide, by contrast, showed weaker amplitudes and only a weak double-peak feature, likely because the first event had already scoured and smoothed the runout path, reducing the intensity of impact interactions with the ground.

The core of the reconstruction came from inverting the low-frequency seismic signals, filtered between 0.035 and 0.1 hertz, into a force-time function. The method treats the sliding mass as an equivalent single-point force acting on the ground and retrieves its temporal evolution by deconvolving the recorded signal with Green’s functions computed from the Crust1.0 global one-dimensional velocity model. Integrating the resulting force divided by the estimated mass once yields centroid velocity, and twice yields displacement. Because the two events’ individual volumes could not be separated from post-event topography, the team used the roughly 2:1 ratio of long-period signal peak amplitudes to apportion the total mass, estimating approximately 4.2 × 10⁸ kilograms for the first landslide and 2.1 × 10⁸ kilograms for the second. A sensitivity analysis spanning mass ratios from 1:1 to 3:1 bounded the uncertainty in the derived kinematics.

The inverted kinematics revealed a three-stage movement process for each event: initiation and acceleration, collision and redirection, and deceleration and accumulation. The first landslide accelerated eastward under gravity, its centroid reaching an initial peak velocity of about 45 meters per second after 19 seconds and roughly 240 meters of displacement. The motion azimuth then swung abruptly from about 90 degrees to 310 degrees as the mass collided with a slope transition, before dropping to 135 degrees at a local topographic rise, where the mass re-accelerated to a second peak of 24 meters per second. It finally ground to a halt in a southeast-oriented gully at about 61 seconds, having traveled approximately 870 meters. The second landslide followed a similar script in compressed time, peaking first at 35 meters per second, deflecting across a topographic scarp, reaching a second peak of 43 meters per second, and stopping after 41 seconds and about 830 meters of total displacement. The authors caution that these velocities and displacements are approximate centroid estimates subject to significant uncertainty from mass estimation, not deterministic values.

A telling detail emerged from the friction analysis. Calculating the sliding friction coefficient during the stable initiation stage, the team found it decreased from about 0.43 to 0.35 in the first event, an 18.6 percent reduction, and from about 0.41 to 0.33 in the second, a 19.5 percent reduction. This systematic weakening during acceleration plausibly reflects the progressive breakdown of asperity interlocking along the sliding surface, fragmentation of the sliding mass into granular material, and enhanced sliding and rolling among blocks, possibly compounded by transient pore-pressure effects that reduced effective normal stress. Notably, an empirical relationship between friction coefficient and landslide volume predicts a coefficient of roughly 0.4 for a volume of 180,000 cubic meters, closely matching the back-calculated values and lending independent credibility to the inversion.

To explain why the slope failed in the first place, the team built a two-dimensional discrete element model in UDEC based on the reconstructed pre-sliding topography, treating the slope as a deformable sliding mass over a rigid bed and progressively reducing shear strength parameters in a strength reduction procedure. The simulation showed deformation concentrating first at the rear edge and toe of the sliding body: tensile cracking developed behind, shear deformation accumulated at the toe, and the two zones progressively coalesced along the potential sliding surface until the mass detached and accelerated downslope. Laboratory uniaxial compression tests on mudstone from the source area, which yielded strengths of 120 to 137 megapascals and an average elastic modulus of about 12.25 gigapascals, confirmed the intact rock’s hard-brittle character, though the authors emphasize that joints, weathering, and water-induced weakening, not intact rock strength, governed the slope-scale behavior.

The geological setting supplied the vulnerability. The Feixianguan Formation’s interbedded soft silty mudstone and hard sandstone predispose slopes to sliding; the mudstone softens in water and weathers spheroidally, while oxidation of pyrite within the formation generates expansive stress that continuously propagates fractures. Seasonal monsoon rainfall, averaging about 1,100 millimeters annually, drives wet-dry cycling that alternately raises pore pressures and shrinks mudstone into tension cracks, and winter freeze-thaw action in shallow fractures further degrades the rock mass. The area also sits within the Huayingshan seismic zone, the most active tectonic zone in the Sichuan Basin, with 57 earthquakes of magnitude 3.0 or greater recorded within 20 kilometers of the site between 2015 and 2024, repeatedly fracturing the rock and opening pathways for water. In the week before failure, roughly 27 millimeters of rain fell, which the authors interpret as a possible short-term contributing factor rather than a sole trigger. Nearby coal mining, 1.2 kilometers from the slide, remains an unquantified potential influence for lack of monitoring data.

Finally, a limit equilibrium sensitivity analysis identified which factors actually controlled stability. Raising the rear tension crack connectivity ratio from 10 to 90 percent slashed the calculated safety factor from 1.4 to 0.54, a 61 percent reduction, because a connected crack simultaneously destroys tensile resistance and channels water pressure into the slope. Increasing the sliding surface’s internal friction angle raised the safety factor by roughly 23 percent across the tested range. By contrast, horizontal seismic acceleration and the water depth in the rear crack had comparatively modest effects, with the safety factor declining only about 4 percent even as crack water depth reached 40 meters. The message is sobering and actionable: for slopes like Junlian’s, it is the slow, invisible deterioration of rear-edge cracks and frictional strength, accumulated over years of wetting, drying, freezing, and shaking, that determines whether a mountainside holds or collapses in under a minute. The seismic-reconstruction framework demonstrated here offers a way to read that story after the fact, and potentially, as regional seismic networks improve, to catch the next one in progress.

Subject of Research: Failure mechanism and dynamics of the February 2025 Junlian landslides in Southwestern China reconstructed through seismic signal inversion, discrete element simulation, and slope stability analysis

Article Title: Failure mechanism analysis of the catastrophic Junlian landslide in Southwestern China based on seismic signal analysis and numerical simulation

Article References: Chen, Q., Tang, Q., Fan, G., Zhou, J., Li, C., & Zhang, B. (2026). Failure mechanism analysis of the catastrophic Junlian landslide in Southwestern China based on seismic signal analysis and numerical simulation. Environmental Earth Sciences, 85(16), Article 409. https://doi.org/10.1007/s12665-026-13142-6

Image Credits: AI Generated

DOI: 10.1007/s12665-026-13142-6

Keywords: Junlian landslide, landslide dynamics, seismic signal analysis, force-time function inversion, discrete element simulation, limit equilibrium analysis, slope stability, Sichuan, high-speed long-runout landslide, Hilbert-Huang Transform, friction coefficient, geohazard reconstruction