Negative sample selection is a major source of uncertainty in landslide susceptibility assessment (LSA), because areas without recorded landslides cannot be directly regarded as stable. Small baseline subset interferometric synthetic aperture radar (SBAS-InSAR) deformation can provide useful constraints for identifying low-deformation candidate areas. However, in steep alpine canyon terrain, low line-of-sight (LOS) deformation may result from unfavorable SAR viewing geometry rather than true slope stability. To address this problem, this study proposes a geometry-aware InSAR feedback purification sampling strategy (GIFPS) for negative sample selection. GIFPS integrates ascending and descending SBAS-InSAR deformation, C-index-based LOS geometric sensitivity, and model feedback to select more reliable negative samples. The method has been evaluated in the Shigatse region of the Qinghai–Tibet Plateau and compared with buffer-controlled sampling (BCS) and dual-orbit low-deformation intersection sampling (DOLIS). Repeated experiments using support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost) show that GIFPS consistently improves model performance. Compared with BCS and DOLIS, GIFPS increases the mean ROC-AUC by 4.77 percentage points and 3.58 percentage points, respectively, and increases the mean F1-score by 3.75 percentage points and 2.59 percentage points, respectively. The ROC-AUC improvements are statistically significant according to paired Wilcoxon signed-rank tests. Susceptibility zoning, SHAP interpretation, and sample-distribution diagnostics further show that GIFPS improves spatial discrimination mainly through reliability-oriented negative sample selection, rather than by excessively narrowing the conditioning-factor distribution of candidate samples. These results suggest that GIFPS provides an interpretable InSAR-assisted strategy for negative sample selection in LSA in complex alpine canyon areas.

In landslide-prone areas, spatial gaps in InSAR-derived deformation maps caused by incomplete SAR coverage hinder continuous surface deformation assessment and limit reliable landslide analysis. To address this problem, we propose an explainable AI (XAI) framework that integrates SBAS-InSAR, ensemble machine learning, and Shapley Additive exPlanations (SHAP) to estimate surface deformation in SAR-scarce regions. Geological and engineering factors, including protective measures, distance to roads, and land use, were combined with remote sensing and field data to build a comprehensive dataset. Four ensemble models (LightGBM, XGBoost, Random Forest, and CatBoost) were trained and evaluated, with XGBoost achieving the best performance (R2 = 0.816, RMSE = 6.85 mm, MAE = 4.27 mm). Validation against two GNSS benchmarks confirmed sub-millimeter accuracy (0.6 mm and 0.3 mm). Both XGBoost and CatBoost delineated continuous deformation patterns consistent with field-observed damage. SHAP analysis provided model interpretability, highlighting elevation and human-engineering factors as key drivers: areas farther from roads and under cultivation were more prone to downslope movement, while damaged protective works exhibited greater deformation. By coupling InSAR with XAI, this study achieves accurate and interpretable surface deformation estimation in data-scarce regions, advancing landslide assessment and early warning applications.

With the operation of large-scale hydraulic infrastructures,many slopes located in reservoir areas are prone to deformation and failure under the influence of fluctuations in the reservoir water level,thereby seriously threatening the safe operation of hydraulic facilities and possibly leading to more severe secondary disasters.Therefore,conducting accurate landslide susceptibility assessment is of great significance for landslide risk prevention and mitigation in reservoir regions.In this study,a refined dynamic landslide susceptibility assessment was conducted for the Zigui-Badong reservoir bank section of the Three Gorges Reservoir area.First,a comprehensive landslide inventory was established on the basis of long-term field investigations and historical records.Subsequently,16 conditioning factors,including elevation,slope,and lithology,were selected to construct a landslide susceptibility evaluation index system,and an initial static susceptibility assessment was performed using ensemble machine learning methods.On this basis,long-term surface deformation information was retrieved using the small baseline subset interferometric synthetic aperture radar(SBAS-InSAR)technique.An optimization matrix that integrates susceptibility classes and SBAS-derived deformation rates was developed to couple static predictions with dynamic deformation information,thereby obtaining dynamic landslide susceptibility results.(1)Very high-and high-susceptibility zones are mainly distributed along the Yangtze River and its tributaries,and elevation,vegetation coverage,distance to rivers,and rainfall are the dominant controlling factors for landslide spatial distribution;(2)compared with individual base models,the ensemble learning model more effectively integrates their advantages and achieves the highest prediction accuracy(AUC=0.954);(3)the dynamic susceptibility results coupled with time-series InSAR data can effectively correct false-negative and false-positive errors in static susceptibility assessments.This study improves the accuracy and timeliness of landslide susceptibility modeling and provides a valuable reference for the dynamic management of landslide hazards in reservoir areas.

Against the backdrop of the global transition toward clean energy, China's Yangtze River Basin has established the world's largest clean energy corridor. However, the stable operation of the Mega Clean Energy Transmission Network (MCETN) in this region is increasingly threatened by landslides under extreme climate conditions. Given the current lack of clarity regarding the extent of landslide impacts on the MCETN, it is critical to systematically assess the potential spatial distribution probability of landslide occurrence across historical, current (integrating historical periods), and future scenarios. To address this, space-time clustering analysis is used to identify time-aggregation windows in landslide inventories, revealing path-dependent effects in historical landslide occurrences. Next, by integrating multi-temporal environmental factor data and applying two ensemble learning frameworks, the spatial distribution of landslide susceptibility within each temporal window was predicted. Finally, leveraging historical landslide data, future landslide susceptibility (2030-2100) under two climate scenarios (SSP2-4.5 and SSP5-8.5) is assessed. The results demonstrate that landslide events exhibit the strongest clustering within an 8-year window. The blending ensemble framework consistently demonstrates optimal performance across all periods, with annual maximum rainfall contributing most significantly to landslide susceptibility modelling, confirming its role as a primary triggering factor. Integrating landslide susceptibility from different historical periods reveals that 6.7% of the study area falls within a very high level. Interestingly, projections under both climate scenarios indicate that a larger proportion of areas within the MCETN will experience an increase in landslide susceptibility index, highlighting the urgent need to enhance infrastructure resilience against escalating climate extremes.

Landslides are a serious geologic hazard common to many countries around the world. They can result in fatalities and the destruction of infrastructure, buildings, roads, and electrical equipment. Especially rapid-moving landslides, which occur suddenly and travel at high speeds for miles, can pose a serious threat to life and property. Landslide inventories are essential to understand the evolution of landscapes, and to ascertain landslide susceptibility and hazard, and it can be of help for any further hazard and risk analysis. Although many landslides inventories have already been created worldwide, often these archives of historical landslide events lack precise information on the date of landslide occurrence. Many of these inventories also lack completeness especially in case of smaller landslides which is also caused by landslides erosion processes, human impact, and vegetation regrowth. Precise determination of landslide occurrence time is a big challenge in landslide research. Optical and Synthetic Aperture Radar (SAR) images with multi-spectral and textural features, multi-temporal revisit rates, and large area coverage provide opportunities for landslide detection and mapping. Landslide-prone regions are frequently obscured by cloud cover, limiting the utility of optical imagery. The capacity of SAR sensors to penetrate clouds allows the use of SAR satellite data to provide a more precise temporal characterization of the occurrence of landslides on a regional scale. The archived Copernicus Sentinel-1 satellite, which has a 6 to 12-day revisit period and covers the majority of the world's landmass, allows for more precise identification of landslide failure timings. The time-series of SAR amplitude, interferometric coherence, and polarimetric features (alpha and entropy) have strong responses to landslide failures in vegetated regions. This is characterized by a sudden increase or decrease in their values. Consequently, the abrupt shifts in the time-series of SAR-derived parameters, triggered by the failure, can be recognized and regarded as the failure occurrence time. The aim of this study is to determine the time period of failure occurrences by automatically detecting abrupt changes in the time series of SAR-derived parameters. We present a strategy for anomaly detection in time-series based on deep-learning to identify the failure time using four parameters derived from SAR time series. In this strategy, we introduce a gated relative position bias to an unsupervised Transformer model to detect anomalies in a multivariate time-series composed of four SAR-derived parameters. We conduct an experiment involving multiple landslides and compare the performance of our proposed strategy for detection of the failure time period with that of the LSTM model. Our strategy successfully identifies the time of landslide failure, which closely approximates the actual time of occurrence when compared to the LSTM model employed in this study.

0 INTRODUCTION Synthetic Aperture Radar(SAR)remote sensing,particularly with the C-band Sentinel-1 mission,has been widely used for landslide displacement analysis due to its high spatial resolution and revisit frequency(Zhou et al.,2024;Dai et al.,2021).However,in densely vegetated or humid mountainous regions such as the Three Gorges Reservoir(TGR),C-band signals suffer from temporal decorrelation,limiting their effectiveness for landslide monitoring.

Regional-scale landslide early warning systems are commonly developed based on empirical rainfall thresholds. However, current rainfall threshold models overlook the significant spatial variations in slopes response to rainfall, thus undermining the accuracy of warnings. In this study, we improved a regional-scale landslide early warning method based on rainfall thresholds by accounting for the varying rainfall sensitivity of individual slope units. First, the Rainfall Sensitivity Index (RSI) for both dry and wet seasons are computed using a Graph Attention Network (GAN) model, incorporating ten influencing factors. Subsequently, an initial regional critical rainfall threshold was obtained based on the Effective cumulative rainfall - rainfall Duration (E-D). Finally, the improved rainfall thresholds for each slope units are derived by coupling the initial critical rainfall threshold and RSI. Using Pingyang County, Zhejiang Province, China as test site, the reliability and reasonableness of the proposed method was validated by statistical analysis, in-situ tests and historical rainfall events. The results reveal that the GAN demonstrates robust predictive capability in RSI assessment, achieving a precision of 0.88. Notably, slopes exhibit higher rainfall sensitivity during dry season compared to wet season. The early warning system employing improved critical thresholds shows significant improvement, with 11.9 % higher accuracy and 14.3 % fewer missed alarms relative to conventional methods. Overall, this study proposes a novel method for downscaling of regional-scale thresholds to slope-unit levels in landslide early warning systems, which informs risk mitigation strategies and governmental decision-making, thereby effectively reducing the risk of landslides.

The prevalent catalog-based Landslide Susceptibility Modelling (LSM) operates under the assumption that future landslide occurrences mirror past and current patterns. Due to growing urban expansion and climate change, certain landslides follow new patterns of occurrence, disrupting the foundational assumption of catalog-based LSM and leading to constraints in the effectiveness of traditional susceptibility maps. Here, to address this problem, we proposed a method to produce more accurate and dynamic landslide susceptibility maps by coupling advanced Ensemble Machine Learning (EML) and Multi-Temporal Interferometric SAR (MT-InSAR). The Wanzhou District in Three Gorges Reservoir area of China is considered as the test site. The landslide catalog and multiple EML methods are used for the preparation of the preliminary susceptibility map. We have also compared and analyzed the impact of ensemble strategies (homogeneous and heterogeneous ensemble) and base-learners on the modelling performance. Subsequently, Sentinel-1 data from 2018 to 2020, analyzed using MT-InSAR approach, are used to map ground deformation rates. We outline the active slopes and deduce the relationship between the deformation of Matou landslide and triggering factors. The final susceptibility map is generated by coupling catalog-based susceptibility and ground deformation rate maps through an empirical assessment matrix. Our results show that the causal factors of distance to rivers, distance to faults, annual rainfall and distance to roads are basic parameters for landslide spatial development; Heterogeneous EML methods outperform the homogeneous, and the more base-learner types provide better performance. InSAR-acquired deformation rates corrected overestimation and underestimation errors in the landslide susceptibility map produced by catalog-based method. Our proposed method is capable of improving the accuracy and timeliness of susceptibility map, providing a useful instrument to better assess landslide risk scenarios in rapidly changing environments.

Since the first impoundment in 2003 of the Three Gorges Reservoir (TGR), one of the largest reservoirs in the world, the issues of slope instability in the Three Gorges Area (TGA) have attracted significant worldwide attention. The operation of TGR, coupled with anthropogenic activities, has influenced slope instability and reactivation of plenty of landslides in the region. This study introduces a methodology to assess the slope instabilities over TGA using advanced integration of hydrological triggering factors with multi-temporal InSAR (MT-InSAR) techniques. Our approach involves characterizing the transient deformation of reservoir bank slopes under the coupling effect of rainfall and reservoir water level (RWL) changes. To achieve this, we propose a methodology that uses MT-InSAR analysis and regression analysis to identify triggering factors, taking into account the periods when slope instability is influenced by the drainage/storage period of the reservoir and seasonal rainfall. Determining the optimal window size for the triggering factors involves iterative searching through wavelet analysis, considering the time-lag between rainfall and RWL data. To extract step-like kinematic features for slowing-moving landslides, we apply a constrained least-squares optimization to InSAR-derived displacement time series. We then use independent component analysis (ICA) to isolate and recover the dominant source features, facilitating unsupervised spatiotemporal clustering to elucidate slope kinematics. This study utilized nearly 100 high-resolution Spotlight TerraSAR-X (TSX) and 50 medium-resolution Sentinel-1 (S1) SAR images captured between 2019 and 2021 to assess the slope instability. Here, we first test our proposed approaches for the single Huangtupo landslide in the TGA, which is one of China's largest reservoir-wading landslides along the Yangtze River; then, the approaches have been expanded to the whole study region near the Badong County. Overall, our proposed framework is transferable and can be applied to other local landslide or regional studies for monitoring slope instability and analyzing complicated cascading hazard chains.

>INTRODUCTION On May 1st,2024,around 2:10 a.m.,a catastrophic collapse occurred along the Meilong Expressway near Meizhou City,Guangdong Province,China,at coordinates 24° 29′24″N and 116° 40′25″E.This collapse resulted in a pavement failure of approximately 17.9 m in length and covering an area of about 184.3 m 2 （Chinanews,2024）.

High-intensity mining activities caused by the rapid increase of coal consumption can lead to unprecedented multi-hazard effects of underground coal mining. Revealing the long-term surface deformation of a coal mining area plays an important role in understanding a disaster. However, due to the limitation of methods and data, the long-term evolution effect of surface deformation in coal mining regions remains unclear. Here, we improved the Interferometric Synthetic Aperture Radar (InSAR) method to process multi-source radar remote sensing monitoring data from the past twenty years to uncover the long-term surface deformation throughout the study region. We found that the initial scattered smaller deformation areas detected by Generic Atmospheric Correction Online Service (GACOS)-assisted Stacking gradually merged into a coherent larger region. Our Improved Interferometric Point Target Analysis (IPTA)-InSAR method showed that coal mining activities can lead to significant surface deformation that can last for several years. The velocity-involved stability assessment method assesses the stability in a mining area using historical time series, so as to judge the current state of this area. Results revealed that GACOS-assisted Stacking is more suitable for extracting surface deformation from coal mining activities at the regional scale as it greatly reduces topographic and atmospheric errors during the data processing. Additionally, the correction effect of GACOS datasets on C-band datasets is better. Meanwhile, we discuss the limitations of proposed stability assessment method. This study provides a new understanding of surface deformation caused by coal mining activities.

The prediction of landslide deformation is crucial for early warning systems. While conventional geotechnical in-situ monitoring is restricted due to its high cost and spatial limitations over large regions, deep learning-based methodologies with remote sensing data have become increasingly prevalent in contemporary predictive research, yet this frequently engenders the enigmatic “black box” issue. To address this, we improve the landslide displacement prediction framework by combining interpretable deep learning based on an attention mechanism and Multi-Temporal Interferometric Synthetic Aperture Radar (MT-InSAR) techniques. MT-InSAR is first used to extract a landslide displacement time series from Copernicus Sentinel-1 SAR images. Then Variational Mode Decomposition (VMD) is employed to separate the nonlinear displacement time series into trend, seasonal, and noise components. The Auto-Regressive Integrated Moving Average (ARIMA) model and Bidirectional Gated Recurrent Unit (BiGRU) are applied to predict trend and seasonal displacements, respectively. The inputs for these predictions are determined by analyzing landslide influencing factors. This study uses the Xinpu landslide in the Three Gorges Reservoir Area of China to evaluate the proposed method and compare its performance with existing models. The CNN-Attention-BiGRU algorithm effectively captures the nonlinear relationship between landslide deformation and its triggering factors, outperforming conventional deep learning models such as BiLSTM, BiGRU, and CNN-BiGRU, achieving improvements in Root Mean Square Errors (RMSEs) by 21%—55% and Mean Absolute Errors (MAEs) by 23%—56%. By applying deep learning with an attention mechanism, our proposed method considers the underlying principles of landslide deformation, and factors with higher relative importance for prediction modeling are interpreted to be concentrated annually between April and August, enabling a more effective and more accurate prediction of large-scale landslide kinematics for the studied reservoir region.

Landslide hazard assessment (LHA) evaluates both the spatial and temporal probability of landslide occurrences, playing a crucial role in risk mitigation and prevention. Recently, LHA is widely applied owing to the accumulated availability of high-quality landslide samples over decades. However, two critical issues remain to be tackled: (1) In a well-established landslide inventory, the time span of their occurrence is recorded to be up to several decades. The critical analysis of the timeframe scale effects of landslide samples (year vs. decade vs. half-century scales) on LHA remains absent in current methodologies. (2) The relationship between landslide occurrence and environmental factors is characterized by complex nonlinearities, making it challenging for a single model to undertake sophisticated modeling tasks. To address these issues, we conducted a study in Hubei Province, China. We collected landslide samples spanning 47 years from 1975 to 2021. Twelve scenarios were designed based on different timeframes of landslide occurrences. Subsequently, we utilized two metaheuristic algorithms, namely the White Shark Optimizer (WSO) and the Chameleon Swarm Algorithm (CSA), to optimize the convolutional neural network (CNN) model for landslide susceptibility assessment. The rainfall thresholds for Hubei Province were derived based on the Rd-Rp model. We calculated the landslide hazard by coupling these thresholds with landslide susceptibility levels through a landslide hazard matrix. Our findings indicate that a 15-year timeframe for landslide samples is most suitable for LHA in Hubei Province, yielding the highest AUC value of 0.873 in this scenario. Furthermore, the CSA algorithm outperforms the WSO in finding the optimal parameters. The LHA based on the proposed method in this study effectively predicted rainfall-induced landslides on July 9, 2024. This demonstrates the practical application and predictive power of our proposed method in landslide hazard assessment.

Quantifying landslide susceptibility saves lives, especially in populous areas exposed to wet climates. However, available hydrological data sets such as precipitation and soil moisture are usually from reanalysis with a few to tens of kilometers' coarse resolution compared to the dimensions of landslides. Here we aim to seek substitutes to characterize hydrological features with finer spacing for landslide susceptibility assessment encompassing the tectonically active California. We synergize remote sensing big data and derivatives including topographic characteristics, vegetation index, hydrological variables, land cover, and geological units in different machine learning architectures. Our results illuminate that the interferometric coherence derived from synthetic aperture radar (SAR) can be an effective hydrological proxy, providing enhanced resolution by three orders of magnitude to tens of meters and presenting satisfactory performance, with recalls >85 % and AUCs >90 % in our landslide susceptibility models. The consequent spatially continuous landslide susceptibility map further demonstrates the effectiveness of high-resolution SAR products in compensating for limitations in traditional hydrological data sets. The map and our inferred relationship with the melange and the distance to faults improve our ability in landslide hazard mitigation.

Objective To investigate the spatial-temporal variations in landslide susceptibility due to human engineering activities in resettled urban areas. Methods This study focuses on the new urban area of Yunyang County in the Three Gorges Reservoir region. Landslide susceptibility time-varying index factors were introduced to map spatial-temporal susceptibility differences and explore the spatial-temporal evolution of landslide disasters during urbanization in resettled urban areas. First, the stacking ensemble model was selected as the static susceptibility evaluation model. Then, the InSAR deformation rates and land use types over three distinct time spans (namely, January 16, 2017, to August 27, 2018 (T1), September 20, 2018, to July 30, 2021 (T2), and August 23, 2021, to November 17, 2023 (T3)) were selected as time-varying factors. Last, the time-varying factors were combined with the static evaluation results to create susceptibility difference distribution maps for the different periods. Results The study revealed that introducing time-varying factors in the analysis of spatial-temporal susceptibility differences effectively reflects the impact of urbanization on landslide disasters. When the land type in the study area changed from non-engineering land to engineering land, the landslide susceptibility level generally increased, with grid shares of 61.3% and 67.1% in the two change stages, respectively. The temporal trends of the InSAR displacement time series curves for selected typical landslides in urban areas showed high spatial-temporal correlations with land type changes, further validating the reliability of this method. Conclusion The proposed research approach provides the basis for disaster prevention, mitigation, and regional planning during the urbanization process in resettled urban areas of the Three Gorges Reservoir region.

The prediction of landslide deformation is an important part of landslide early warning systems. Displacement prediction based on geotechnical in-situ monitoring performs well, but its high costs and spatial limitations hinder frequent use within large areas. Here, we propose a novel physically-based and cost-effective landslide displacement prediction framework using the combination of Multi-Temporal Interferometric Synthetic Aperture Radar (MT-InSAR) and machine learning techniques. We first extract displacement time series for the landslide from spaceborne Copernicus Sentinel-1A SAR imagery by MT-InSAR. Using wavelet transform, we then decompose the nonlinear displacement time series into trend terms, periodic terms, and noises. The advanced machine learning method of Gated Recurrent Units (GRU) is utilized to predict the trend and periodic displacements, respectively. The modeling inputs for trend and periodic displacement predictions are determined by analyzing their corresponding influencing factors. The total displacements are finally predicted by summing the predicted displacements of trend and periodic items. The Shuping and Muyubao landslides, identified as seepagedriven and buoyancy-driven, respectively, in the Three Gorges Reservoir area in China are selected as case studies to evaluate the performance of our methodology. The prediction results demonstrate that machine learning algorithms can accurately establish the nonlinear relationship between the landslide deformation and its triggers. GRU outperforms the algorithms of Long Short-Term Memory networks and Kernel-based Extreme Learning Machine, and the Adam algorithm can effectively optimize the model hyperparameters. The root mean square error and mean absolute percentage error are 3.817 and 0.022 in Shuping landslide, and 5.145 and 0.020 in Muyubao landslide, respectively. By integrating the advantages of MT-InSAR and machine learning techniques, our proposed prediction framework, considering the physics principles behind landslide deformation, can predict landslide displacement cost-effectively within large areas.

Landslide susceptibility evaluation is pivotal for mitigating landslide risk and enhancing early warning systems. Current practices in developing Landslide Susceptibility Mapping (LSM) often overlook the diverse mechanisms of landslides, and traditional machine learning (ML) models lack the capability for autonomous feature learning in landslide contexts. This study proposes a methodology that precedes the application of deep learning algorithms for LSM by classifying landslides and selecting relevant factors based on their deformation mechanisms. In the Zigui-Badong section of the Three Gorges Reservoir area (TGRA), landslides are classified into rock landslides (RL) and soil landslides (SL) based on the geological conditions and historical landslide inventory. A comprehensive evaluation index system, comprising thirteen factors is established. To identify the most pertinent factors for each type of landslide, these factors are ranked according to their contribution to landslide occurrence. For susceptibility assessment, this study introduces a Convolutional Neural Network (CNN) model and benchmarks its performance to traditional ML models including Classification and Regression Trees (CART) and Multilayer Perceptrons (MLP). The efficacy of these models is evaluated using the Receiver Operating Characteristic (ROC) curve and various statistical analysis methods. The findings indicate that LSMs that consider different types of landslides yield more accurate and realistic outcomes. The CNN model outperformes its counterparts, with MLP being the second most effective and CART the least effective. Overall, this study demonstrates the superiority of an LSM approach that accounts for landslide diversity over traditional, monolithic methods.

Determining the timing of landslide occurrence is crucial for establishing an accurate, comprehensive and systematic landslide inventory while assessing the potential for reducing landslide risk. Unfortunately, many existing landslide inventories lack temporal information such as the precise time of landslide events. Optical and Synthetic Aperture Radar (SAR) sensors are the most commonly used remote sensing technologies for landslide detection. Unlike optical sensors, SAR sensors are not affected by cloudy conditions and provide valuable imagery regardless of sunlight availability. Therefore, SAR-derived parameters, i.e., SAR amplitude, interferometric coherence, and polarimetric features (alpha and entropy), offer a higher temporal resolution for detecting landslide occurrence times compared to optical data. Despite the advantages, there is currently no universally accepted automatic method for determining the time of landslide events using SAR data. This is due to the lack of anomaly labels and the high time-series volatility in detecting landslide occurrence times. Despite advances in deep-learning methods for anomaly detection in time-series, only a few of them can address these challenges in our case. In this paper, we propose an unsupervised multivariate transformed-based deep-learning model to automatically and efficiently estimate landslide occurrence times using multivariate SAR-derived parameters time-series analysis. The designed gated relative position can increase robustness and temporal context information, by learning global temporal trends in the time-series. Subsequently, the time-series of the anomaly score derived from the proposed Transformer model is analyzed using an adaptive thresholding strategy to dynamically and automatically mark anomalies related to the landslide occurrence. Our research focuses on collapsed landslides characterized by dramatic changes in ground surface topography, with a particular attention for the need of a prior knowledge about landslide boundaries. We assess the performance of the proposed methodology for several collapsed landslides including the July 21, 2020 Shaziba and 23 July, 2019 Shuicheng landslides in China, March 19, 2019 Takht landslide in Iran, June 15, 2018 Jalgyz-Jangak and May 25, 2018 Kugart landslides in Kyrgyzstan, July 7, 2018 Hitardalur landslide in Iceland, and January 25, 2019 Brumadinho landslide in Brazil. In comparison to commonly used neural networks like the LSTM algorithm, our proposed framework leads to a more accurate estimate for the time of landslide failure using time-series of SAR-derived parameters. Furthermore, our results suggest the great potential of SAR data to narrow the time period detected from optical data when used in conjunction with them.

This paper investigates the spatiotemporal characteristics and life-cycle of movements within the Joshimath landslide-prone slope over the period from 2015 to 2024, utilizing multi-sensor interferometric data from Sentinel‑1, ALOS‑2, and TerraSAR‑X satellites. Multi-temporal InSAR analysis before the 2023 slope destabilization crisis, when the region experienced significant ground deformation acceleration, revealed two distinct deformation clusters within the eastern and middle parts of the slope. These active deformation regions have been creeping up to −200 mm/yr. Slope deformation analysis indicates that the entire Joshimath landslide-prone slope can be categorized kinematically as either Extremely-Slow (ES) or Very-Slow (VS) moving slope, with the eastern cluster mainly exhibiting ES movements, while the middle cluster showing VS movements. Two episodes of significant acceleration occurred on August 21, 2019 and November 2, 2021, with the rate of slope deformation increasing by 20

In recent years, several catastrophic landslide events have been observed throughout the globe, threatening to lives and infrastructures. To minimize the impact of landslides, the need of landslide susceptibility map is important. The study aims to extract high-quality non-landslide samples and improve the accuracy of landslide susceptibility modelling (LSM) outcomes by applying a coupled method of ensemble learning and Machine Learning (ML). The Zigui-Badong section of the Three Gorges Reservoir area (TGRA) in China was considered in the present study. Twelve influencing factors were selected as inputs for LSM, and the relationship between each causal factor and landslide spatial development was quantitatively analyzed. A total of 179 landslides have been used in the present study. About 70% of the landslide pixels were randomly considered for training, and the remaining 30% were used for validation. Logistic Regression (LR) model was applied to produce an initial susceptibility map, and the non-landslide samples were selected within the classified low-susceptibility zone. Subsequently, two ML classifiers - the Classification and Regression Tree (CART), and the Multi-Layer Perceptron (MLP), and four coupling models - the CART-Bagging, CART-Boosting, MLP-Bagging, and MLP-Boosting, were utilized for LSM. Finally, the receiver operating characteristics (ROC) curve and statistical analysis were applied for accuracy assessment. The results show that altitude and distance to rivers were the main causal factors of landslides in the study area. The LR-MLP-Boosting performed the best with an accuracy of 0.986 followed by the LR-CART-Bagging, LR-CART-Boosting, and LR-MLP-Bagging. Accuracy comparisons demonstrate that ensemble learning algorithm can notably enhance the LSM performance of ML classifiers, and the Boosting algorithm marginally outperforms the Bagging algorithm. Moreover, the LR model can effectively constrain the selection range of non-landslide samples. The non-landslide sampling method constrained by LR yields higher quality samples compared to raditional random sampling method with no constraints, which develops a more excellent LSM.