Search

In this work, we present the CAS Landslide Dataset, a large-scale and multisensor dataset for deep learning-based landslide detection, developed by the Artificial Intelligence Group at the Institute of Mountain Hazards and Environment, Chinese Academy of Sciences (CAS). The dataset aims to address the challenges encountered in landslide recognition. With the increase in landslide occurrences due to climate change and earthquakes, there is a growing need for a precise and comprehensive dataset to support fast and efficient landslide recognition. In contrast to existing datasets with dataset size, coverage, sensor type and resolution limitations, the CAS Landslide Dataset comprises 20,865 images, integrating satellite and unmanned aerial vehicle data from nine regions. To ensure reliability and applicability, we establish a robust methodology to evaluate the dataset quality. We propose the use of the Landslide Dataset as a benchmark for the construction of landslide identification models and to facilitate the development of deep learning techniques. Researchers can leverage this dataset to obtain enhanced prediction, monitoring, and analysis capabilities, thereby advancing automated landslide detection., (© 2024. The Author(s).)

An increasing number of dumpsites have been built to stockpile a large amount of construction solid waste (CSW) during urbanization. Physical vulnerability assessment to CSW landslides plays a critical role in evaluating the safety risk of the potentially affected buildings. A catastrophic CSW landslide occurred in Shenzhen, China, in December 2015, which directly destroyed 75 buildings along the travel path. In this study, based on the statistical fundamental information of the damaged buildings and the reproduced motion process of the landslide using numerical modeling, both simulation-based progressive destruction process characterization and statistical-based analyses of the relationships between intensity data and vulnerability indexes are comprehensively performed to quantify the physical vulnerability of the damaged buildings, identify the failure models, and determine critical factors for controlling the building damages subjected to the Shenzhen CSW landslide. A typical vertical continuous collapse model due to the bending failure of the bottom column and the upward extension of the plastic hinge area is observed in the established single-frame structure under a set of impact pressure data. The occurrence of building failures close to the time of peak impact pressures controlled by the dynamic pressure indicated by velocity, along with the recorded flow velocity and impact pressure data exceeding the corresponding threshold values for all affected buildings, both suggest that flow velocity and impact pressure indicators are more appropriate for quantifying the vulnerability of buildings exposed to high-mobility CSW landslides. The conclusions could help CSW dumpsite stakeholders to better perform risk assessment and mitigate hazard propagation. [ABSTRACT FROM AUTHOR]

The conventional numerical solvers for partial differential equations encounter a formidable challenge, as their computational efficiency and accuracy are heavily contingent on grid size. Recently, machine learning (ML) has exhibited substantial promise in addressing partial differential equations. Nevertheless, substantial hurdles persist in practical applications. In this work, we endeavor to establish a deep learning framework founded on the Fourier neural operator (FNO) for resolving the intricacies of simulating real landslide dynamic processes. Our findings demonstrate that the current FNO approach adeptly replicates landslide dynamic processes and boasts exceptional computational efficiency. Additionally, it is noteworthy that this data‐driven ML methodology can seamlessly incorporate data from other experimental sources or numerical simulation techniques. Consequently, this work underscores the significant potential of utilizing ML methodologies to supplant conventional numerical simulation methods. Plain Language Summary: There are great challenges involved in leveraging machine learning methods to learn realistic physical dynamic processes. When it comes to the real landslide movement across intricate terrains, it is meaningful to validate the capacities of machine learning in tackling the complicated problem. This study aims to propose an innovative solution of modeling of landslide dynamic processes from a machine learning perspective. Here, we introduce a data‐driven framework based on Fourier neural operator to predict the dynamic behavior of actual landslides. Following an exhaustive assessment, the superior performance of our suggested model in real landslide situations and its versatility in adapting to landslides across various geographical regions have been confirmed. This study explores a new approach to modeling landslide dynamic processes and highlights the great potential of data‐driven approaches to address dynamic process challenges present in real physical world. Key Points: The data‐driven deep learning method based on FNO can achieve fast prediction of real landslide accumulation processThe proposed data‐driven method for predicting landslide dynamic processes can be extended to new areas after transfer learningWe provide numerical datasets of landslide dynamics, which can serve as the foundational resources for ML‐based landslide forecasting tasks [ABSTRACT FROM AUTHOR]

The efficiency and accuracy of grid-based computational fluid dynamics methods are strongly dependent on the chosen cell size. The computational time increases exponentially with decreasing cell size. Therefore, a grid coarsing technology without apparent precision loss is essential for various numerical modeling methods. In this article, a physical adaption neural network (PANN) is proposed to optimize coarse grid representation from a fine grid. A new convolutional neural network is constructed to achieve a significant reduction in computational cost while maintaining a relatively accurate solution. An application to numerical modeling of dynamic processes in landslides is firstly carried out, and better results are obtained compared to the baseline method. More applications in various flood scenarios in mountainous areas are then analyzed. It is demonstrated that the proposed PANN downscaling method outperforms other currently widely used downscaling methods. The code is publicly available and can be applied broadly. Computing by PANN is hundreds of times more efficient, meaning that it is significant for the numerical modeling of various complicated Earth-surface flows and their applications. [ABSTRACT FROM AUTHOR]

Numerical prediction of landslide runout and deposition is important for estimating landslide risk and developing mitigation plans. The choice of a suitable model and its parameters and a confident calibration strategy are crucial for numerical simulations. Here, we evaluated two constitutive models with a three‐dimensional smoothed particle hydrodynamics (SPH) method by simulating the catastrophic 11 October 2018 Baige landslide. The results indicate that both the soil mechanic and fluid models can capture the dynamic runout and deposition morphology while using different values of input parameters. A point‐wise comparison of deposit elevation can minimize the calibration error. Numerical models were constrained accurately by utilizing both the static observation data and dynamic seismic signals. The effects of friction on deep‐seated landslides motion and deposition are more significant than cohesion. The 3D model includes the effects of shear stresses and velocities inside the material body, resulting in a reduced friction coefficient compared to the 2D model (e.g., depth‐averaged model). Our study highlights the potential of the 3D SPH method for modeling large‐scale complex landslides. Plain Language Summary: Landslides belong to a type of earth surface process recognized by their high damage potential. Computer models can simulate the landslides' movement to predict speed, forces, and deposition, which can help to delineate areas at risk and to design mitigation measures. We simulated the 2018 Baige landslide in China with a method known as smoothed particle hydrodynamics (SPH), to test whether it is suitable for landslide simulation and to investigate protocols for model calibration when including seismic data in addition to information on the landslide's deposit. In this method, the landslide body is represented by millions of small, moving, interacting particles, which offers advantages over established models in terms of computation time and details in the simulation. We find that SPH is suitable for modeling large‐scale natural landslides. The seismic data are more valuable in the model calibration than landslide deposit observations. The new approach yields landslide simulations that deliver more details on particle velocities within the landslide body, and their spatial and temporal distribution. These details can be used to infer the properties of landslides during the sliding process, which helps to better understand landslides in general and to set up models for events where little data are available. Key Points: A three‐dimensional mesh‐less graphics processing unit‐accelerated smoothed particle hydrodynamics formulation is suitable for modeling large‐scale natural landslidesThe selection of an appropriate constitutive model depends on the landslide features and materialsIncluding dynamic seismic signals into numerical model calibration routines in addition to topographic changes improves calibration [ABSTRACT FROM AUTHOR]

Assessing the hazard of potential landslides is crucial for developing mitigation strategies for landslide disasters. However, accurate assessment of landslide hazard is limited by the lack of landslide inventory maps and difficulty in determining landslide run-out distance. To address these issues, this study developed a novel method combining the InSAR technique with a depth-integrated model. Within this new framework, potential landslides are identified through InSAR and their potential impact areas are subsequently estimated using the depth-integrated model. To evaluate its capability, the proposed method was applied to a landslide event that occurred on November 3, 2018 in Baige village, Tibet, China. The simulated results show that the area with a probability of more than 50% to be affected by landslides matched the real trimlines of the landslide and that the accuracy of the proposed method reached 85.65%. Furthermore, the main deposit characteristics, such as the location of maximum deposit thickness and the main deposit area, could be captured by the proposed method. Potential landslides in the Baige region were also identified and evaluated. The results indicate that in the event of landslides, the collapsed mass has a high probability to block the Jinsha River. It is therefore necessary to implement field monitoring and prepare hazard mitigation strategies in advance. This study provides new insights for regional-scale landslide hazard management and further contributes to the implementation of landslide risk assessment and reduction activities. [ABSTRACT FROM AUTHOR]

Automatic recognition and segmentation methods have become an essential requirement in identifying large-scale earthquake-induced landslides. This used to be conducted through pixel-based or object-oriented methods. However, these methods fail to develop an accurate, rapid, and cross-scene solution for earthquake-induced landslide recognition because of the massive amount of remote sensing data and variations in different earthquake scenarios. To fill this research gap, this paper proposes a robust deep transfer learning scheme for high precision and fast recognition of regional landslides. First, a Multi-scale Feature Fusion regime with an Encoder-decoder Network (MFFENet) is proposed to extract and fuse the multi-scale features of objects in remote sensing images, in which a novel and practical Adaptive Triangle Fork (ATF) Module is designed to integrate the useful features across different scales effectively. Second, an Adversarial Domain Adaptation Network (ADANet) is developed to perform different seismic landslide recognition tasks, and a multi-level output space adaptation scheme is proposed to enhance the adaptability of the segmentation model. Experimental results on standard remote sensing datasets demonstrate the effectiveness of MFFENet and ADANet. Finally, a comprehensive and general scheme is proposed for earthquake-induced landslide recognition, which integrates image features extracted from MFFENet and ADANet with the side information including landslide geologic features, bi-temporal changing features, and spatial analysis. The proposed scheme is applied in two earthquake-induced landslides in Jiuzhaigou (China) and Hokkaido (Japan), using available pre- and post-earthquake remote sensing images. These experiments show that the proposed scheme presents a state-of-the-art performance in regional landslide identification and performs stably and robustly in different seismic landslide recognition tasks. Our proposed framework demonstrates a competitive performance for high-precision, high-efficiency, and cross-scene recognition of earthquake disasters, which may serve as a new starting point for the application of deep learning and transfer learning methods in earthquake-induced landslide recognition. [ABSTRACT FROM AUTHOR]

In tectonically active steep mountain belts, such as those in west Sichuan in China, frequent landslide disasters can occur during the rainy season. High-locality landslides often initiate with the failure of fractured rock masses or regolith near the ridge of a high and steep slope. Often, these events occur suddenly in the sense that people downslope are caught by surprise. High-locality landslides can be destructive owing to the large momentum and entrainment potential of the moving mass. Increasing attention has been given to these events since the 2008 Wenchuan earthquake. Indeed, slope ridges are particularly susceptible of topographic amplification of the seismic shaking, resulting in mechanical weakening that predisposes them to enhanced weathering and rain-triggered failures years or decades after the mainshock. Risk reduction in such cases must begin with a systematic surveying of the ridges to identify criticalities. Then, quantitative (numerical) approaches must be used to evaluate volumes and describe the movement. Risk assessments can then be produced, and mitigations suggested. Here, we describe the Tonghua landslide in Li County, which is currently in a creeping stage but threatens critical infrastructures and dwellings. We monitored the site for 3 years and identified additional potentially unstable areas before performing physically based modeling and hazard zonation. Our results could be useful in decision-making for risk reduction. [ABSTRACT FROM AUTHOR]

Copyright of Canadian Geotechnical Journal is the property of Canadian Science Publishing and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In this study, based on the frictional velocity-weakening law used in continuum modeling, an adapted Hertz-Mindlin contact model between particles and the ground surface is established to accurately simulate the landslide dynamics process. Moreover, the long-period seismic waves indicating the dynamic characteristics of large landslides are adopted to determine the computational parameters for the discrete element method (DEM). By combining the force-time functions obtained from the long-period seismic waves and numerical modeling, the dynamic parameters and reasonableness of the simulation results of the first Baige landslide on October 11, 2018, in eastern Tibet, China, have been validated. The friction coefficient obtained in this study is also in agreement with the relationship between the volume and friction coefficients reported in the literature. Taking the same model and parameters, the dynamic process leading to the second Baige landslide on November 3, 2018, is analyzed. The modeled profile and position of the maximum thickness of the deposits arising from the two landslides both match the observations well. It is demonstrated that this combination procedure is feasible for the prediction and identification of areas susceptible to large landslides. [ABSTRACT FROM AUTHOR]

Landslide disasters are associated with severe losses on the Loess Plateau of China. Although early warning systems and susceptibility mapping have mitigated this issue to some extent, most methods are qualitative or semi-quantitative in the site-specific range. In this paper, a quantitative spatial distribution model is presented for site-specific loess landslide hazard assessment. Coupled with multi-temporal remote sensing images and high-precision UAV cloud point data, a total of 98 loess landslides that have occurred since 2004 on the Heifangtai terrace were collected to establish a landslide volume-date and retreating distance database. Eleven loess landslides are selected to construct a numerical model for parameter back analysis, and the accuracy of the simulation results is quantitatively evaluated by the centroid distance and overlapping area. Different volumes and receding distance rates of landslides are fitted to determine the relationship between cracks and potential volume, and different volumes and parameters are combined to simulate the spatial distribution of potential loess landslides. The results of this study reveal that landslide volumes mainly range between 1 × 103 and 5 × 105 m3, and the historical occurrence probability reaches 0.551. The optimal parameters are estimated by the maximum likelihood method to obtain a uniform distribution parameter value probability model, and the results show that the error of the estimated length within a range of 0.05 from the optimal parameter does not exceed 15%. In the selected slope slide case, farmland near the toe of the slope primarily includes exposed hazards with probabilities greater than 0.7. This work provides a useful reference for local disaster reduction and a theoretical methodology for hazard assessments. [ABSTRACT FROM AUTHOR]

A consequent slope, comprising weak planes in the same dip direction along the slope face, is the most dangerous type of rock slope. Herein, a consequent landslide that occurred on May 5, 2018, in Mabian County, Southwest China, was investigated. This landslide occurred in a region with a complex geological background and strong geological activity and is considered a typical consequent landslide. Field investigations and aerial photographic interpretations revealed that the rock mass slipped along the contact surface of the strongly and moderately weathered rock mass, triggered by continuous rainfall. A depth-integrated continuum method was used to study the dynamic process of the landslide. The applicability of the depth-integrated method was verified by comparing the moving boundary of the landslide extracted from a video with the simulation results. The results show that the velocity of the landslide first increased and then decreased from the top to the bottom of the landslide and reached a maximum of 21 m/s in the middle of the landslide. Based on this method, we predicted the potential danger zones and velocity distribution of unstable masses next to the Mabian landslide. On comparing the calculated results of a depth-integrated continuum method with the Scheidegger method, in the case of the Mabian landslide, a reduction factor λ was added to the Scheidegger formula to achieve a modified formula that determines the velocity of such consequent landslides relatively accurately. This work could provide reference for the study of landslide velocity and disaster prevention. [ABSTRACT FROM AUTHOR]

Successive major landslides during October and November 2018 in Baige village, eastern Tibet, dammed the Jinsha River on two occasions, and the subsequent dam breaches instigated a multi-hazard chain that flooded many towns downstream. Analysis of high-resolution aerial images and field investigations unveiled three potentially unstable rock mass clusters in the source area of the landslides, suggesting possible future failures with potential for river-damming and flooding. In order to evaluate and understand the disaster chain effect linked to the potentially unstable rock mass, we systematically studied the multi-hazard scenarios through an integrated numerical modelling approach. Our model begins with an evaluation of the probability of landslide failure, including runout and river damming, and then addresses the dam breach and resultant flood—hence simulating and visualising an entire disaster chain. The model parameters were calibrated using empirical data from the two Baige landslides. Then, we predict the future cascading hazards via seven scenarios according to all possible combinations of potential rock mass failure. For each scenario, the landslide runouts, dam-breaching, and flooding are numerically simulated with full consideration of uncertainties among the model input parameters. The maximum dam breach flood extent, depth, velocity, and peak arrival time are predicted at sequential sites downstream. As a first attempt to simulate the full spectrum of a landslide-induced multi-hazard chain, our study provides insights and substantiates the value provided by multi-hazard modelling. The integrated approach described here can be applied to similar landslide-induced chains of hazards in other regions. [ABSTRACT FROM AUTHOR]

High-locality landslides primarily occur in the upper zone of the failed slope, in which the slip surface is considerably higher than the foot of the slope. This paper discusses a high-locality landslide that occurred on August 8, 2017, on the left bank of Zagunao River in Tonghua Town, southwest China. The landslide, called the Tonghua landslide, originated on a 716-m-high hillslope, which was formed by preexisting glacial deposits. The main body of the landslide is currently suspended on hillslope, with a relative relief of 353 m and is undergoing progressive failure. The town of Tonghua is located downslope of this landslide, and hence, the landslide could cause severe damage, including casualties and the destruction of infrastructure. Very few mitigation measures have been implemented to reduce the potentially devastating consequences due to the high locality and continuous movement of the landslide. We conducted comprehensive studies including field investigations, aerial photographic interpretations, pluviometric regime analysis, and superficial displacement monitoring to understand the failure mechanism and evolutionary process of this ongoing landslide. Considering that a Ms 7.0 earthquake occurred in Jiuzhaigou, approximately 220 km northeast of Tonghua Town, on the same day as the landslide, and that sporadic but persistent rainfall occurred before slope failure, we believe that both the rainfall and the earthquake were potential triggers for this landslide. However, further studies are necessary to better understand the characteristics and to develop hazard mitigation strategies for this type of landslide. [ABSTRACT FROM AUTHOR]

The focus of this paper is on two large sequential landslides which formed a dam and the resulting lake that occurred along the Jinsha River on October 11 and November 3, 2018. About 24 and 9 × 106 m3 of material failed and rushed into the river, respectively. Both landslides totally blocked the river and induced hazard chains. Four multi-temporal unmanned aerial vehicle (UAV)-based high-resolution images and digital elevation models (DEMs) before and after the landslides were obtained and merged. The initial deformation of the two landslides can be clearly captured by optical remote-sensing images and interferometric synthetic aperture radar (InSAR) analysis. The depth-integrated continuum method was adopted to analyze the dynamic process of both landslides, using the same basal friction angle and cohesion. The computational results matched well with field measurements and observations. It is shown that the evaluation of potentially landslide-prone areas by the depth-integrated continuum method is feasible and is able to provide significant information before a possible event. [ABSTRACT FROM AUTHOR]

Ridge-top rockslides frequently cause huge property losses and casualties due to the difficulties involved in detecting their precursors by means of manual surveys. Their early identification and the surrounding area's evaluation in terms of potential danger are essential for preventing disasters. The recent large rockslide which occurred at Su Village, which is in possession of data related to pre-failure images, real-time video, and post-failure boreholes, is helpful for providing new insights into the processes associated with these events. Due to a strong sustained rainfall, a large volume of granite blocks rapidly descended from the upper part of the hillside, causing 27 deaths and destroying more than 20 houses. It is found that the early identification of the hazard associated with such rockfalls can be made by the analysis of remote sensing images and Persistent Scatterer Interferometry (PSI) analysis of Synthetic Aperture Radar (SAR) data. The depth-integrated continuum method, including taking the entrainment effect into account, was adopted to analyze the dynamic processes and to identify the areas at risk. The computational results show that the evaluated runout distance and extent match well with the field investigation results. The parameter sensitivity surrounding cohesion, coefficients of lateral earth pressure, and volume amplification were analyzed. It is demonstrated that the cohesion plays a significant role in the dynamic processes and the deposited area. However, the effects from the earth pressure coefficient and volume bulking are comparatively weaker. [ABSTRACT FROM AUTHOR]

Construction solid waste (CSW) landfill landslides, such as the Guangming New District landslide, which occurred in Shenzhen (hereafter the Shenzhen landslide), occur when the material is loose and saturated. They usually exhibit characteristics such as abrupt failure and whole collapse. During the propagation of landslides, dilatation behavior plays an important role in causing liquefaction, resulting in high velocity and exceptionally long run-out dynamics. We propose a dynamic model for describing fluidized CSW landslides by integrating the dilatancy model into smoothed particle hydrodynamics (SPH). The dilatancy model implies that the occurrence of dilation or the contraction of the granular-fluid mixture depends on the initial solid volume fraction. The dynamic model is used to simulate the Shenzhen landslide, and special attention is paid to the effects of different initial solid volumes on the mobility of the CSW landslide. The results show that when the solid volume fraction is higher than the critical value, contraction occurs, the excess pore water pressure increases, and the basal friction resistance is reduced. CSW landslide mobility is based on the initial solid volume fraction (or initial void ratio) of the granular-fluid mixture; a slight change in the initial volume fraction significantly affects the mobility of the CSW landfill landslide. [ABSTRACT FROM AUTHOR]

Rock avalanches represent a serious risk for human lives, properties, and infrastructures. On June 24, 2017, a catastrophic landslide destroyed the village of Xinmo (Maoxian County, Sichuan, China) causing a large number of fatalities. Adjacent to the landslide source area, further potentially unstable masses were identified. Among them, a 4.5-million m3 body, displaced during the landslide event by about 40 m, raised serious concerns. Field monitoring and a reliable secondary risk assessment are fundamental to protect the infrastructure and the population still living in the valley. In this framework, the use of distinct element methods and continuum model methods to simulate the avalanche process was discussed. Various models (PFC, MatDEM, MassMov2D, Massflow) were used with the aim of reproducing the Xinmo landslide and, as predictive tools, simulating the kinematics and runout of the potentially unstable mass, which could cause a new catastrophic event. The models were all able to reproduce the first-order characteristics of the landslide kinematics and the morphology of the deposit, but with computational times differing by several orders of magnitude. More variability of the results was obtained from the simulations of the potential secondary failure. However, all models agreed that the new landslide could invest several still-inhabited buildings and block the course of the river again. Comparison and discussion of the performances and usability of the models could prove useful towards the enforcement of physically based (and multi-model) risk assessments and mitigation countermeasures. [ABSTRACT FROM AUTHOR]

Since lots of underground and slope excavation works were conducted during the urbanization process, an increasing number of sites in ravines around a city have been used to stockpile a large amount of excavated soils. This brings a huge challenge for researchers and managers in the risk evaluation and mitigation of potential dangers of these man-made construction waste landfills. This paper describes a recently large landslide of the construction waste landfill, which occurred at a site of Guangming new district in Shenzhen, China, on December 20, 2015. This catastrophic landslide caused the death of 69 persons and 8 persons are still missing. In this paper, this landslide was numerically simulated and analyzed. In spite of neither high-intensity rainfall nor antecedent rainfall, a slope of this landfill with a relative height of 111 m sided and caused about 2.34 million cubic meters of the soils to travel over a gentle terrain more than 1.2 km. This means that the landslide mobility index (H/L = 0.092) is much lower than a general designed value and the values in most other cases. A depth-integrated continuum method and a MacCormack-TVD finite difference algorithm are adopted, in this paper, to numerically simulate the dynamic process of this large landslide. It is found that a Coulomb friction model with consideration of the pore water pressure effects can well reproduce the main characteristics of the dynamic process of this landslide. Sensitivity analysis has demonstrated that the high pore water pressure in the soils plays a significant role in its mobility and is a key factor to the severity of this landslide. [ABSTRACT FROM AUTHOR]

This study used a three-dimensional model based on the finite-discrete element method-smoothed particle hydrodynamics (FDEM-SPH) coupling approach to reconstruct the 11 October 2018 Baige landslide that blocked the Jinsha river. The numerical model simulated the dynamic process of the landslide by FDEM, while the SPH simulated the behavior of the river water. The landslide deposit area and the water-eroded area in the FDEM-SPH simulation agree well with the results of the field investigation. According to the simulation, the main duration of the Baige landslide was 80 s. The peak average speed of the landslide was 34.4 m/s, while the local speed of the sliding mass reached a maximum velocity of 70 m/s. The tensile effect and the shear effect caused isotropic dispersive stress which caused the dynamic and violent fragmentation of the landslide. The model also simulated the variation of kinetic energy, the accumulated friction dissipation, and the fracture energy of the sliding mass with time. After the sliding mass ran into the river, the landslide triggered a wave, which rapidly dispersed due to the subsequent mass movement. The front part of the sliding mass pushed some of the river water to the opposite bank where the wave reached a height of 120 m. The results of the FDEM-SPH model, the particle flow code (PFC) model, and the depth-integrated shallow-water flow (DISWF) model are compared and analyzed concerning studying landslides that block rivers. Three types of evolution mechanisms of high-level flow-like landslide-induced waves in deep river valleys are proposed in this paper. • A novel fluid-solid coupling approach based on FDEM-SPH is used for landslide blocking river. • Dynamic process of the Baige landslide blocking river is shown clearly. • Evolution modes of high-level fluid-like landslides blocking rivers in deep valley area are proposed. [ABSTRACT FROM AUTHOR]