Application of Artificial Intelligence and Remote Sensing for Landslide Detection and Prediction: Systematic Review
Abstract
:1. Introduction
- Over the years, what are the commonly used AI, ML, and DL methods for landslide hazard mapping?
- What are the commonly used remote sensing techniques in landslide hazard studies?
- How accurate are the AI, ML, and DL, as well as the remote sensing techniques, capable of evaluating landslides?
2. Materials and Methods
2.1. Literature Search, Exclusion, and Inclusion Strategy
- The study focuses on landslide susceptibility mapping, detection, prediction, and deformation monitoring with remote sensing data, imagery, satellite imagery or data, AI, ML, and DL, and no other type of landslide study.
- The articles are written only in English.
- The research articles were published in reliable journals.
2.2. Data Extraction
2.3. Data Analysis
3. Results
3.1. Literature Search Characteristics
3.2. Geographic Distribution and Research Publication Trends
3.3. Major Topics in the Reviewed Research Article
3.4. Commonly Used Remote Sensors
3.5. Learning Algorithms with Remote Sensor Techniques
4. Remote Sensing and Learning Algorithms Application for Landslides
4.1. A Brief Overview of the Application of Learning Models and Remote Sensing
4.2. Landslide Mapping with AI, DL, ML, and Remote Sensing Data
4.2.1. Landslide Susceptibility Mapping
- (1)
- Marjanovic et al. [13] assessed the landslide susceptibility in Fruška Gora Mountain, Serbia, using remote sensing imagery derived from the digital elevation model (DEM); landslide environmental factors such as elevation, aspect, lithology, topographic wetness index; and three different machine learning algorithms, support vector machine (SVM), decision tree (DT), and logistics regression (LR). The results from unbalanced and balanced training modeling, the area under the curve (AUC) for SVM, provide the best results for landslide susceptibility mapping.
- (2)
- Stumpf et al. [83] combined the object orientation image analysis with varieties of remote sensing datasets obtained from Geoeye-1, IKONOS, and Quickbird and obtained landslide susceptibility mapping accuracies ranging from 73% to 87% for four different landslide sites in China, France, Haiti, and Italy.
- (3)
- Peethambaran et al. [23] compared the fuzzy expert system (FES) and extreme learning machine (ELM) using remote sensing images from Alos Pulsar DEM (15 m resolution), satellite images (LISS-III and Radarsat), and eight causative factors of several landslides in Uttarakhand, India. It was found that the FES outperformed the ELM with an AUC of 0.84 compared to 0.80.
- (4)
- Qi et al. [17] analyzed 14,397 shallow landslides in Southern Tianshui, China, using 18 geomorphic concentration factors, including slope aspect, geological conditions, and human activities. Their results indicated that out of 21 ML models developed, the ExtraTrees model produced better outcomes, with an AUC of 0.91. It was also found that the slope terrain aspect was the most significant contributor to landslide susceptibility compared to the other factors.
- (5)
- Yin et al. [79] explored remote sensing data from the Qinghai-Tibet Plateau to map common landslides based on environmental factors such as terrain and climate. The ML demonstrated that climate conditions, such as summer temperature and rainfall, were the main contributing factors causing landslides.
- (6)
- Zeng and Chen [74] investigated an unstable slope using DEM-derived data and established artificial neural networks (ANNs) and GIS-based techniques with relevant environmental factors to map landslide susceptibility in Enshi, China. Their proposed models successfully predicted the unstable slope location based on AI theory and GIS. Later, Zeng et al. [136] proposed a new model for environmental heterogeneity for Fengjie County and Fulin District in Chongqing, China, which was based on a graph neural network with environmental consistency (GNN–EC) and compared it with three different learning models such as SVM, ANN, and random forest (RF). The investigators reported a landslide susceptibility accuracy higher than 0.80, a precision of 0.80, a recall of 0.91, an F-measure of 0.81, and an AUC of 0.87, indicating a strong capability of mapping landslide susceptibility compared to the other three learning models.
- (7)
- An innovative susceptibility-guided (SG) landslide-mapping method based on a fully convolutional neural network (FCNN) and mean changing the magnitude of objects (MCMO) was proposed by Chen et al. [137] to improve the landslide hazard mapping in Lantau Island, Hong Kong, China. This method produced better outcomes with a precision, recall, and F1-score between 80% and 95% and an intersection over union (IoU) value between 70% and 85%, indicating a significant reduction in false and landslide detections compared to the traditional FCNN.
- (8)
- Garcia et al. [92] proposed a semi-automatic detection of relict landslides in the Serra do mountain range, Southeastern Brazil, based on a deep learning CNN framework (Unet, FPN, and Linknet). The k-means cluster algorithm was used for pre-training, and weights were computed to fine-tune the training process in the CNN. Forty-two CNN tests of two datasets were performed, achieving recall values higher than 75% and a precision of less than 20%. It was found that the model predicted relict landslides accurately but was limited to detecting landslides in the terrains covered with rainforests.
- (9)
- Zhang et al. [138] proposed a prototype-guided domain-aware progressive representation learning (PG-DPRL) technique to map cross-domain landslides based on the multitarget domain adaptation (MTDA) method. The model adopted a near-to-far adaptation strategy, reliable domain-specific pseudo-label, and cross-domain sharing boundary decision. A category-wise representative alignment was performed to enhance the discriminative capacity of the model by using the Wasserstein distance metric and cross-domain prototype consistency loss. Using landslide datasets from Hokkaido, Japan, the study demonstrated that cross-domain landslides could be successfully mapped using the PG-DPRL model and global high-resolution landslide datasets, and the proposed model outperformed many traditional deep learning algorithms in comparison.
- (10)
- Ouyang et al. [139] developed a model (PU-pullbggingDT) algorithm to cater for the limitations of the positive unlabeled (PU) algorithm in mapping landslide susceptibility and prediction. The model uses the global shuttle radar topography mission (SRTM), Google Earth Engine (GEE) datasets, and 25 environmental factors associated with landslides in Zigui County, China. The proposed PU-pullbggingDT model outperformed the existing SVM, DT, LR, AdaBoost, XGBoost, and PU-learning, indicating a strong capability of accurately predicting landslides in complex geological terrains.
- (11)
- Zhao et al. [20] combined the CNN and transformer model to develop CNN–transformer local–global extraction network features (CTLGNet) to investigate landslide hazards in the Three Gorges Reservoir and Jiuzhaigou areas, China. Considering nine landslide environmental factors, the CTLGNet performed better than other methods such as the CNN, residual neural network (ResNet), densely connected convolutional network (DenseNet), vision transformer (ViT), and fractional Fourier image transformer (FrIT).
4.2.2. Landslide Detection
- (1)
- Ghorbanzadeh et al. [32] used high-resolution optical satellite data from RapidEye, 20 landslide environmental condition factors, and maps combined with the the ANN, different CNN, SVM, and RF models to detect landslide-hazard-prone areas along the Ganga River in the Himalayas, northern India. This study showed that the CNN model achieved a mean intersection over union (mIoU) of 0.782 with insignificant window size among other models, indicating the model’s ability to detect landslides.
- (2)
- Similarly, Hu et al. [54] utilized a cross-validation method and nighttime imagery, Landsat-8, ASTER DEM, and Visible Infrared Imaging Radiometer Suite (VIIRS) from the disaster-prone Jiuzhaigou region with ANN, RF, and SVM considering 30 landslide condition factors. The investigators noticed that the precision and accuracy of landslide identification were reduced for the three models, especially for ANN and SVM. They attributed the accuracy reduction to the “users” and “producers” imbalance accuracy. However, the models can be improved to detect landslides through the cross-application probability.
- (3)
- Sameen et al. [140] designed a residual network (ResNet) for landslide detection in the Cameron Highlands, Malaysia, utilizing altitude, slope, aspect, and curvature. The DEM and light detection and ranging (LiDAR) were used to create spatial data. The ResNet was compared to traditional deep learning models like CNN, and the results showed that the CNN model enhanced landslide detection with the fusion method. However, the ResNet model outperformed the CNN with the same fusion method and showed convergence to the test area with an overall training and validation accuracy and F1-score between 0.80 and 0. 95 and an mIoU of 90.24%.
- (4)
- Nhu et al. [12] adopted LR, logistic model tree (LMT), and RF models to analyze 152 landslide sites in the Cameron Highlands, Malaysia, utilizing Google Earth Images (GEI), interferometry synthetic aperture radar (InSAR), DEM, and field datasets. Seventeen landslide environmental condition factors were considered to detect landslides using the machine learning algorithms with different accuracy measurement statistical tools, including sensitivity, AUC, accuracy, and root-mean-square error (RMSE). This study demonstrated an AUC of 0.92 for LMT, 0.90 for LR, and 0.88 for RF, showing that the datasets from remote sensors and learning algorithms such as LMT could detect landslides.
- (5)
- Cai et al. [160] reported that the DenseNet model reached a Kappa value of about 0.95 and an F1-score of 0.95 after the data augmentation and fine-tuning technique, indicating the model’s applicability to landslide detection. This research around the Three Gorges Reservoir, China, utilized datasets from the ZY-3 satellite, global digital elevation (GDEM), and Landsat-8 with 12 topographic, geological, hydrological, and land cover factors. However, it was noted that the model had poor micro-landslide detection and needed improvement.
- (6)
- Huang et al. [161] demonstrated that incorporating Landsat-8 OLI, advanced spaceborne thermal emission and reflection radiometer, and global digital elevation model (ASTER GDEM) data with nine geospatial landslide factors could benefit a landslide detection analysis. The investigators developed a Distilled Swin–Transformer (DST) model with Swin–Transformer as a backbone to eliminate the long runtime and insufficient model challenges facing ML and DL. They tested this model to detect landslides in Zigui County, Hubei Province, China. In a comparison of the quantitative results with ResNet, data-efficient image transformers (DeiT), and the proposed DST model, the DST model achieved a higher landslide detection accuracy of 98.17%, a precision of 98.17%, a recall of 98.16%, an F1-score of 98%, and a Kappa index of 0.977, with less floating-point operation (FLOP) for landslide detection.
- (7)
- Li et al. [88] indicated that an image-based data-driven framework with satellite images could benefit landslide detection. The proposed approach comprised two essential phases: firstly, an object detection algorithm, thus (Faster-RNN) trained within large-scale satellite imagery and proposed bounding box for each landslide visualization; and secondly, the bounding box location information was utilized to crop the satellite images and boundary detection algorithms proposed to identify each loess landslide dissection performance. A total of 150 loess landslides from northern China were reviewed to validate the effectiveness of the two frameworks. The results showed that the frameworks could accurately detect the segmentation of loess landslides.
- (8)
- To eliminate the anomalous deformation area in InSAR data and human interactions with acceptable criteria, Zhang et al. [73] developed two-stage deep learning networks (InSARNet). The model was utilized to identify anomalous deformation terrains in Maoxian County, China. From the quantitative analysis with different measurement metrics, the InSARNet model performed better than the commonly used learning algorithms in detecting landslide anomalous deformation areas. However, the model was recommended to have future improvements.
- (9)
- Sreelakshmi et al. [162] tackled the challenges of detecting landslides in rainforest vegetation in Bijie City, China, using deep-learning frameworks with visual saliency and high-resolution remote sensing imagery. The saliency feature maps were employed to fine-tune the U-Net model on the landslide data. The results showed that the proposed model achieved a landslide detection accuracy of 94% compared to the commonly used deep-learning algorithms.
4.2.3. Prediction of Landslide Offset
- (1)
- Tien et al. [64] employed a hybrid intelligence model of the least square support vector machine (LSSVM) and artificial bee colony (ABC) optimization called LSSVM–BC to study landslides in the Lao Cai area, Vietnam. Using DEM, 10 landslide environmental influence factors, and 340 landslides, the computational and quantitative analysis showed that the hybrid intelligent model LSSVM–BC could predict landslide offset relatively well compared to the SVM learning algorithm.
- (2)
- Van Natijne et al. [162] used remote sensing data and an ML algorithm to predict landslide deformation. This study investigated the deep-seated Vogelsberg landslide near Innsbruck, Austria, using a long short-term memory (LSTM) model. The investigators pointed out that the LSTM model produced a positive outcome for landslide deformation prediction; however, the ML technique was rather complex.
- (3)
- Chen et al. [164] developed a neural disaster emergency ontology (NADE) schematic layer based on a constructed knowledge graph and different environmental factors, including geology, landform, soil, climate, vegetation, and transportation. The proposed knowledge graph embedding (KGE)-based model was applied to generate landslide prediction in Xiji County, Ningxia Province, China. The model was trained with 741 landslide records from DEM and SRTM. The findings demonstrated an F1-score improvement of 5% with the complete data and 17% with the reserve data.
- (4)
- Zhou et al. [168] combined the physics-based and economical landslide displacement prediction framework with a multi-temporal interferometric synthetic aperture radar (MT-InSAR) and ML method such as gate recurrent units (GRU) to predict the non-linear tendency and time displacement of landslides. In the Three Gorges Reservoir area, China. The GRU model outperformed the traditional ML algorithms, including LSTM, Kernel-based extreme learning machine (KELM), and the Adam algorithm.
- (5)
- He et al. [25] proposed a new method involving photogrammetric and aerial LiDAR surveys and GIS-based kinematic evaluation to analyze the rock slope conditions of the North Coast of Cornwall, UK. The ML models, such as SVM, RF, MLP, and deep learning neural networks (DLNNs), and the discontinuities factors were applied. A frequency analysis (FR) produced better outcomes, suggesting that the inclusion of geological discontinuities into learning algorithms could enhance the prediction accuracy.
- (6)
- Maxwell et al. [169] proposed an explainable boosting machine (EBM) with a generalized additive model (GAM) as an empirical method to better predict landslides in West Virginia, USA. The EBM model was compared with standard ML models such as LR, RF, KNN, and SVM, producing an overall AUC of 0.90, which indicated that the EBM model provided a relatively high accuracy of landslide prediction
- (7)
- Huang et al. [170] considered 11 landslide environmental factors and developed four ML models, LR, MLP, SVM, and C5.0 DT, for Huichang County, China. The C5.0 DT model outperformed others, with an AUC of 0.94.
4.2.4. Landslide Inventory and Monitoring
- (1)
- Lei et al. [171] developed a novel approach to detecting landslides in complex geological terrains and geospatial uncertainty based on a fully convolutional network-fusing pyramid pooling (FCN–PP). The proposed model utilized the Zeiss RMK TOP 15 aerial survey camera system to collect landslide data from Hong Kong, China. The results indicated that the FCN–PP model effectively mapped landslides with a precision of 96%, recall of 96.5%, overall error of 93%, F-score of 90%, and accuracy of 81.90%.
- (2)
- Althuwaynee et al. [172] developed a model incorporating 77 slope deformation factors. They used the t-distribution stochastic neighbor embedding (t-SNE) and the Apriori algorithm to recognize common relationships in the inventory maps with landslide factors. The model was validated using Landsat–TM and provided practical significance for landslide inventory mapping in Pohang state in South Korea.
- (3)
- Ramos-Bernal et al. [173] developed ASTER DEM-derived datasets with primary landslide factors and ML model KNN, stochastic gradient descent (SGD), linear Kernel–SVM, support vector machine radial basis function (SVM–RBF Kernel), and AdaBoost. A total of 671 landslide sites were analyzed, while 2/3 of them were trained, and 1/3 were used to obtain the inventory maps. The study showed that among the five models, AdaBoost recorded a precision, recall, F1-score, accuracy, and Kappa > 90%, with mean errors below 2%, indicating the model’s capacity to provide a landslide inventory map in Guerrero, Southern Mexico.
- (4)
- Chen et al. [174] proposed an algorithm based on multi-feature independent components analysis (MICUNet3+) for landslide inventory mapping with 3 landslide elevation factors. The model was compared with the UNet3+ and Unet3+-RFI using the co-seismic landslide datasets from Jiuzhaigou County, Sichuan Province, China. The MICUNet3+ model had a recall of 0.68, an F1-score of 0.70, and an mIoU of 0.76, indicating that the MICUNet3+ model performed better than the other two models and could assist with detecting landslides in complex geological terrains.
- (5)
- Chen et al. [79] adopted the full convolution networks with focus loss (FCN–FL) ML models such as SVM and RF to map landslides in Bijie City, Guizhou Province, China. The study applied a K-fold cross-validation training model (FCN–FLK) to enhance the data and model sturdiness. The model achieved an accuracy of 0.93, recall of 0.76, F1-score of 0.62, and an mIoU of 0.68, which indicated that it could solve the data imbalance in the landslide inventory mapping. However, the model was limited to the area with large vegetation coverage.
- (6)
- In recent years, some researchers have used remote sensing data and imagery from Sentinel-2, ALOS–PALSAR, DEM, Gaofen-1, and Gaofen-2 with learning paradigms including ANN, RF, CNN, DCN, and LSTM [58,73,91,175]. Sheng et al. [58] studied the implementation of a dynamic approach to landslide-related hazards in Shiyan City, China by integrating spatial–temporal likelihood analysis considering periodic ground deformation velocity derived from the MT-InSAR technique. The study employed ML and statistical models, including information quantity (IQ), FR, LR, BP–ANN, RBF–ANN, RF, SVM, and CNN, and the result indicated that the distance to a river and from structure, slope angle, and rocks were the main factors controlling landslide development. The ML models performed better than the statistical method. From the effectiveness, F1-score, sensitivity, and AUC calculations, the CNN recorded a high value greater than 0.90 and outperformed the other model in landslide monitoring and prediction.
- (7)
- Zhou et al. [176] proposed a model based on deep learning (AtmNet) to deal with the effect of topography and climatic conditions in the landslide-prone terrain of Mao County, Sichuan, China. The landslide-monitoring data were obtained from Sentinel-1 SAR. The data analysis showed that AtmNet provided a platform for DInSAR and InSAR to monitor landslides in mountainous terrains.
- (8)
- Liu et al. [177] utilized an advancing landslide detection method with the multi-period dataset for the landslide-prone area in Tuanjiecun. The small baseline subset interferometric synthetic aperture radar (SBAS–InSAR), GEE, Gaofen-1 (G-1), Gaofen-2 (G-2), Ziyuan-2 (ZY-3), UAV, and DEM were applied for dataset validation. Long short-term memory (LSTM) was used to evaluate and identify the landslide deformation from the SBAS–InSAR quantitatively. The results demonstrated that the LSTM could identify and monitor potential landslide deformations.
- (9)
- Senogles et al. [174] developed a method called “SlideSim” with an optical flow predictor for 3D landslide deformation utilizing DEM and UAV to monitor a slow-moving landslide. The landslide dataset from Southern Oregon Coast, USA was used to validate the model, and the results indicated that the “SlideSim” model required fewer intuitive parameters with no supervision to monitor landslide displacement successfully.
Study Area | Source | Model Evaluation | Reference |
---|---|---|---|
Xi’an, China | Zeiss RMK TOP 15 Aerial Survey Camera System | Full convolutional network within pyramid pooling (FCN–PP) | Lei et al. [171] |
Pohang, South Korea | GEE, Landsat-8, Enhanced thematic mapper (ETM+) | t-distributed stochastic neighbor embedding (t-SNE) | Althuwaynee et al. [172] |
Guerrero, Mexico | ASTER, DEM | KNN, stochastic gradient descent (SGD), support vector machine (SVM linear Kernal), support vector machine radial basis function (SVM RBF Kernel), and AdaBoost | Ramos-Bernal et al. [173] |
Guizhou, China | GEE, Sentinel-2 | Full convolutional networks with focus loss (FCN–FL) | Chen et al. [79] |
Sichuan, China | DEM, Sentinel-2, Landsat-8 | UNet3+, MICUNet3+ and UNet3+-RFI | Chen et al. [174] |
India | Sentinel-2, ALOS–PALSAR, DEM | ANN, RF, and analytical hierarchy process (AHP) | Nath et al. [91] |
South Oregon Coast, USA | DEM, UAV | SlideSim | Senogles et al. [175] |
Maoxian County, China | SBAS–InSAR, SRTM, DEM | InSARNet | Zhang et al. [73] |
Tuanjiecun, China | SBAS–InSAR, GEE, Gaofen-1 (GF-1), Gaofen-2 (GF-2), Ziyuan-3 (ZY-3) and Unmanned Aerial Vehicle (UAV) and DEM | Long short-term memory (LSTM) | Liu et al. [177] |
Shiyan, China | Multi-Temporal InSAR (MT–InSAR) | Information quantity (IQ), FR, LR, BP–ANN, RBF–ANN, RF, SVM, and CNN | Sheng et al. [58] |
Sichuan, Ghana | InSAR, DEM, GEE, SRTM and Sentinel-1 SAR | AtmNet | Zhou et al. [176] |
4.2.5. Assessment of Landslide Potential
- (1)
- Lin et al. [125] assessed landslide susceptibility using two case studies of Taiwan Island with landslide data derived from InSAR. The ML models DT, RF, AdaBoost, XGBoost, and NB considered geological and topographic factors for quantitative analysis. The results revealed that the RF model achieved a better precision accuracy of 82.95%, demonstrating its capability of assessing and predicting landslides in Taiwan Island.
- (2)
- Chang et al. [97] assessed non-landslide data from the Yangtze River in Yichang City, China, along the Three Gorges Reservoir areas prone to landslide. The Landsat-8 OLI, DEM, GEE, and ALOS datasets were analyzed with an unsupervised deep embedding clustering (DEC) algorithm with a deep integration network known as the capsule neural network based on SENet and SE–CapNet. The study results showed that the SE–CapNet had the highest prediction assessment (AUC = 0.97) of landslides, with rainfall being the main driving factor for landslides in the Yangtze River basin.
- (3)
- Mabdeh et al. [127] used two new genetic algorithms (GAs) based on bagging constructed with DT, KNN, and NB models based on bagging and random sub-space (RS) technique to assess and map landslide susceptibility. Considering 16 landslide environmental factors and Landsat-8 and DEM-derived data from Ajloun and Jerash, Jordan, the RS-based ensemble model produced a high accuracy (0.95) in assessing landslide.
- (4)
- Arrogate et al. [179] assessed landslide recognition using supervised and unsupervised models based on explainable ML models, as well as continuous change detection and classification (CCDC) models. With landslide data obtained from GEE, Landsat time series, and SRTM in addition to 16 landslide influence factors, the semi-automatic CCDC model showed a strong capability of assessing landslide risk in Guerrero State in Mexico.
- (5)
- Shao et al. [184] utilized the LR model with remote sensing images from GEE, ALOS–PALSAR, and DEM-derived data, with influencing factors, including elevation, slopes, aspect, lithology, and topography, to assess potential landslide risk in the downstream area of Jinsha River, Tibet Plateau. The data indicated that the LR model effectively identified landslide hazards around Jinsha hydropower station.
- (6)
- Chen et al. [109] assessed landslide susceptibility in the Tumen River Basin, utilizing a principle component analysis (PCA) and 15 landslide influence evaluation factors. The investigators developed an Adaptive Neuro-Fuzzy Inference System (ANFIS) based on Particle Swarm Optimization (PSO), Artificial Bee Colony algorithm (ABC), Shuffled Frog Leaping Algorithm (SFLA), and Bat algorithm (BAT). The ABC model achieved higher accuracy metrics measurements like AUC, root mean square error (RMSE), and mean absolute error (MAE) compared to the other models.
4.2.6. Landslide Extraction and Management
- (1)
- Liu et al. [34] proposed a new model based on the DL algorithm (U-Net model) for automatic extraction landslides. The model considered three RGB bands and landslide-influencing factors to obtain the U-Net + six channels + ResNet model to eliminate the traditional U-Net problem. The Jiuzhaigou County earthquake data from Sichuan Province, China, was used to validate the new model, and the obtained results gave a precision accuracy of 91.3%, recall of 95.4%, and mIoU of 87.5%, compared to the U-Net model.
- (2)
- Using remote sensing images and data from the Gorganroud watershed, Iran, obtained from GEE, ALOS PALSAR, DEM, Sentinal-2, Landsat-8 OLI, and landslide inventory generated with GIS, Arabameri et al. [117] evaluated three ML methods, RF, alternative decision tree (ADTree), and fisher’s linear discriminant function (FLDA). The RF model recorded the highest AUROC of 0.97 and a prediction rate of 0.98, providing important outcomes for managing, predicting, and controlling the current and future landslides.
- (3)
- He et al. [25] employed aerial LiDAR survey, GEE, UAV, and DEM-derived data together with ML and DL models, RF, SVM, MLP, and DLNN to extract and predict landslide potential vulnerabilities. The validated results from the ML models achieved an accuracy of 87% and an AUC of 0.94.
- (4)
- Qi et al. [185] developed an ML algorithm with UAV-sensing image technology. The developed model was capable of extracting landslide features and providing reasonable effects.
- (5)
- Xia et al. [186] derived a landslide extraction model using a full convolution spectral-topographic fusion network (FSTF-Net) based on CNN and geospatial data. The landslide data were obtained from resources satellite-3 and high-resolution remote-sensing technologies such as Beijing-2, SuperView-1, DEM, and Worldview-3 in Mangkam County, Qinghai–Tibet. The proposed FSTF-Net model gave a landslide extraction and recognition precision of 0.85 and an accuracy of 0.89, compared to the current Deeplab_v3+.
- (6)
- Yang et al. [187] developed a background-enhancement technique that could learn the difference between landslide and background features. The landslide influencing factors were added to further enhance the extraction model’s accuracy. The proposed Mask-R-CNN + background-enhancement + landslide influencing factors were applied to the Ludian County landslide in 2014. Using the data from GEE, GeoEye-01, DEM, and Airbus Maxar Technologies, the proposed model recorded an F1-score of 89%, indicating its ability to accurately assess landslide susceptibility, compared to the traditional DL models.
- (7)
- A study conducted by Chen et al. [188] in Lanzhou City, China, introduced a new model using a squeeze-and-excitation network (SENet). This model combined the SENet with U-Net and utilized the data from the remote sensing images. The results indicated landslide extraction effectiveness with an F1-score of 87.94%, compared to the U-Net and U-Net backbone.
5. Conclusions
- Artificial intelligence models ANN, FES, FSP, and XAI; deep learning models CNN, ResU-Net, DCN, and Deeplab V3+; machine learning models XGBoost, RF, LR, and KNN; and several other algorithms based on the traditional learning algorithms have been developed and proposed in the literature for the susceptibility mapping, prediction, and detection of landslides.
- The remote-sensing techniques with high-resolution data sources commonly used in the literature include DEM, ALOS–PALSAR, Landsat OLI, Sentinel groups, InSAR, SRTM, GEI, and other free remote sensors and satellite-available data sources from online. Commercial remote-sensing technologies such as UAV and LiDAR have been popular due to their ability to manage vast areas rapidly. They can also operate at lower heights and can access complex geological terrains.
- The AI, ML, and DL algorithms and remote-sensing technologies reasonably evaluate landslide susceptibility mapping, prediction, detection, monitoring, and inventory with a prediction accuracy ranging from 56% to almost 100%.
- Remote sensing data and imagery have been used with the learning models to study landslide characteristics, especially in susceptibility mapping and detection, followed by assessment and monitoring, with few studies focusing on stability, influence of trained data, and landslide management. Several traditional and newly proposed AI, DL, and ML algorithms available in the literature could be effectively used for landslide hazard assessments and provide local authorities with important information for landslide risk management and control.
- Landsat groups of remote sensing and digital elevation models were widely utilized and found in almost all the research articles reviewed. However, other remote-sensing technologies exist, such as high-resolution unmanned aerial vehicles, sentinel-1 and sentinel-2, advanced spaceborne thermal emission and reflection radiometers, and global digital elevation maps, which better predict and detect landslides.
- Environmental landslide-influencing factors such as topography, geology, and climate were found in almost all the studies reviewed. However, the available learning algorithms and remote-sensing techniques cannot be generalized for predicting landslide hazard studies worldwide, and this deficiency calls for critical attention to provide a generalized and accepted technology and the learning models’ standard range of prediction accuracy metrics.
- Many learning algorithms and models have been developed for a specific case study, which may not apply to other cases. In addition, some models were found to have significant limitations and must be improved.
- There is a lack of studies on landslide susceptibility and learning algorithms, especially in significant parts of Africa, South America, and the Eastern part of Europe. This calls for more research to extend the application of remote-sensing technologies and learning algorithms in assessing landslide susceptibility worldwide.
- Investigating additional remote sensing technologies and learning algorithms for real-time landslide monitoring could help researchers and geoengineers create reliable early warning systems to detect landslides.
Author Contributions
Funding
Acknowledgments
Conflicts of Interest
Abbreviations
Abbreviations | Explanations | Abbreviations | Explanations |
AI | Artificial Intelligence | NB | Naïve Bayes |
DL | Deep Learning | DT | Decision Tree |
ML | Machine Learning | LNRF–BRT | Linear Multivariate Regression–Boosted Regression Tree |
CNN | Convolutional Neural Network | RMSE | Root-Mean-Square Error |
DEM | Digital Elevation Model | XGBoost | Extreme Gradient Boosting |
GEE | Google Earth Engine | KNN | K-Nearest Neighbors |
GEI | Google Earth Image | FR | Frequency Ratio |
ALOS–PALSAR | Advanced Land-Observed Satellite–Phase Array L-Band Synthetic Aperture Radar | LR | Logistics Regression |
LiDAR | Light Detecting Range | FSP | Fuzzy Set Procedure |
InSAR | Interferometric Synthetic Aperture Radar | GNN-EC | Graph Neural Network-Environmental Consistency |
ASTER | Advanced Spaceborne Thermal Emission and Reflection Radiometer | YOLO-SA | You Only Look Once-Small Attention |
UAV | Unmanned Aerial Vehicle | XAI | Explainable Artificial Intelligence |
SRTM | Sensors and Shuttle Radar Topography Mission | FCN–FL | Full Convolutional Network–Focus Loss |
MODIS | Moderate Resolution Imaging Spectroradiometer | ABC | Artificial Bee Colony |
SVM | Supportive Vector Machine | KELM | Kernal-based Extreme Learning Machine |
RF | Random Forest | EBM | Explainable Boosting Machine |
ANN | Artificial Neural Network | ZY-3 | Ziyuan-3 |
AUC | Area Under Curve | G-2 | Gaofen-2 |
mIoU | Mean Intersection Over Union | ANFIS | Adaptive Neuro-Fuzzy Inference System |
CF–RF | Certainty Factor–Random Forest | MAE | Mean Absolute Error |
DCN | Deep Convolutional Network | GB | GradientBoost |
MLP | Multiplayer Perception | TSS | True Skill Statistics |
ELM | Extreme Learning Machine | AUROC | Area Under Receiver Operating Characteristics |
FES | Fuzzy Expert System | ROC | Receiver Operating Characteristics |
DNN–HBA | Deep Neural Network–Honey Badger Algorithm | DLNN | Deep Learning Neural Network |
DADRCNN | Deep Attention Dilated Residual Convolutional Neural Network | GDEM | Global Digital Elevation Model |
LSTM | Long Short-Time Memory | FSTF-Net | Full Convolutional Spectral Topographic Fusion Network-Net |
LD–BiSTM | Landslide Density–Bidirectional Long Short-Term Memory | BPNN | Back-Propagation Neural Network |
GA | Genetic Algorithm | WRF | Weight Random Forest |
BRT | Boost Regression Tree | WLR | Weight Logistic Regression |
CART | Classification And Regression Tree | LSSVM–BC | Least Square Supportive Vector Machine |
GLM | Generalized Additive Model | IoU | Intersection Over Union |
References
- Pradhan, B. A comparative study on the predictive ability of the decision tree, support vector machine and neuro-fuzzy models in landslide susceptibility mapping using GIS. Comput. Geosci. 2013, 51, 350–365. [Google Scholar] [CrossRef]
- Huang, F.; Yin, K.; Huang, J.; Gui, L.; Wang, P. Landslide susceptibility mapping based on self-organizing-map network and extreme learning machine. Eng. Geol. 2017, 223, 11–22. [Google Scholar] [CrossRef]
- Aleotti, P.; Chowdhury, R. Landslide hazard assessment: Summary review and new perspectives. Bull. Eng. Geol. Environ. 1999, 58, 21–44. [Google Scholar] [CrossRef]
- Nadim, F.; Kjekstad, O.; Peduzzi, P.; Herold, C.; Jaedicke, C. Global landslide and avalanche hotspots. Landslides 2006, 3, 159–173. [Google Scholar] [CrossRef]
- Guzzetti, F.; Carrara, A.; Cardinali, M.; Reichenbach, P. Landslide hazard evaluation: A review of current techniques and their application in a multi-scale study, Central Italy. Geomorphology 1999, 31, 181–216. [Google Scholar] [CrossRef]
- Pham, B.T.; Bui, D.; Prakash, I.; Dholakia, M. Evaluation of predictive ability of support vector machines and naive Bayes trees methods for spatial prediction of landslides in Uttarakhand state (India) using GIS. J. Geomat. 2016, 10, 71–79. [Google Scholar]
- Akgun, A.; Türk, N. Landslide susceptibility mapping for Ayvalik (Western Turkey) and its vicinity by multicriteria decision analysis. Environ. Earth Sci. 2009, 61, 595–611. [Google Scholar] [CrossRef]
- Reichenbach, P.; Rossi, M.; Malamud, B.D.; Mihir, M.; Guzzetti, F. A review of statistically-based landslide susceptibility models. Earth-Sci. Rev. 2018, 180, 60–91. [Google Scholar] [CrossRef]
- Kavzoglu, T.; Teke, A.; Yilmaz, E.O. Shared Blocks-Based Ensemble Deep Learning for Shallow Landslide Susceptibility Mapping. Remote Sens. 2021, 13, 4776. [Google Scholar] [CrossRef]
- Li, W.; Fang, Z.; Wang, Y. Stacking ensemble of deep learning methods for landslide susceptibility mapping in the Three Gorges Reservoir area, China. Stoch. Environ. Res. Risk Assess. 2021, 36, 2207–2228. [Google Scholar] [CrossRef]
- Nhu, V.H.; Mohammadi, A.; Shahabi, H.; Ahmad, B.B.; Al-Ansari, N.; Shirzadi, A.; Clague, J.J.; Jaafari, A.; Chen, W.; Nguyen, H. Landslide susceptibility mapping using machine learning algorithms and remote sensing data in a tropical environment. Int. J. Environ. Res. Public Health 2020, 17, 4933. [Google Scholar] [CrossRef] [PubMed]
- Nhu, V.H.; Mohammadi, A.; Shahabi, H.; Ahmad, B.B.; Al-Ansari, N.; Shirzadi, A.; Geertsema, M.; Kress, V.R.; Karimzadeh, S.; Kamran, K.V.; et al. Landslide detection and susceptibility modelling on Cameron highlands (Malaysia): A comparison between random forest, logistic regression and logistic model tree algorithms. Forests 2020, 11, 830. [Google Scholar] [CrossRef]
- Marjanović, M.; Kovačević, M.; Bajat, B.; Voženílek, V. Landslide susceptibility assessment using SVM machine learning algorithm. Eng. Geol. 2011, 123, 225–234. [Google Scholar] [CrossRef]
- Kim, J.C.; Lee, S.; Jung, H.S.; Lee, S. Landslide susceptibility mapping using random forest and boosted tree models in Pyeong-Chang, Korea. Geocarto Int. 2018, 33, 1000–1015. [Google Scholar] [CrossRef]
- Hong, H.; Shahabi, H.; Shirzadi, A.; Chen, W.; Chapi, K.; Ahmad, B.B.; Roodposhti, M.S.; Yari Hesar, A.; Tian, Y.; Tien Bui, D. Landslide susceptibility assessment at the Wuning area, China: A comparison between multi-criteria decision making, bivariate statistical and machine learning methods. Nat. Hazards 2019, 96, 173–212. [Google Scholar] [CrossRef]
- Kwan, J.S.H.; Leung, W.K.; Lo, F.L.C.; Millis, S.; Shi, J.W.Z.; Wong, M.S.; Kwok, C.Y.T. Territory-Wide Identification of Geological Features on Aerial Photographs Using Machine Learning for Slope Safety Management. Springer Ser. Geomech. Geoeng. 2020, 3, 527–539. [Google Scholar]
- Qi, T.; Zhao, Y.; Meng, X.; Chen, G.; Dijkstra, T. Ai-based susceptibility analysis of shallow landslides induced by heavy rainfall in Tianshui, China. Remote Sens. 2021, 13, 1819. [Google Scholar] [CrossRef]
- Ghorbanzadeh, O.; Shahabi, H.; Crivellari, A.; Homayouni, S.; Blaschke, T.; Ghamisi, P. Landslide detection using deep learning and object-based image analysis. Landslides 2022, 19, 929–939. [Google Scholar] [CrossRef]
- Li, Y.; Wang, P.; Feng, Q.; Ji, X.; Jin, D.; Gong, J. Landslide detection based on shipborne images and deep learning models: A case study in the Three Gorges Reservoir Area in China. Landslides 2023, 20, 547–558. [Google Scholar] [CrossRef]
- Zhao, Z.; Chen, T.; Dou, J.; Liu, G.; Plaza, A. Landslide Susceptibility Mapping Considering Landslide Local-global Features Based on CNN and Transformer. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17, 7475–7489. [Google Scholar] [CrossRef]
- Chen, W.; Pourghasemi, H.R.; Naghibi, S.A. A comparative study of landslide susceptibility maps produced using support vector machine with different kernel functions and entropy data mining models in China. Bull. Eng. Geol. Environ. 2018, 77, 647–664. [Google Scholar] [CrossRef]
- Chen, Z.; Zhang, Y.; Ouyang, C.; Zhang, F.; Ma, J. Automated landslides detection for mountain cities using multi-temporal remote sensing imagery. Sensors 2018, 18, 821. [Google Scholar] [CrossRef]
- Peethambaran, B.; Anbalagan, R.; Shihabudheen, K.V.; Goswami, A. Robustness evaluation of fuzzy expert system and extreme learning machine for geographic information system-based landslide susceptibility zonation: A case study from Indian Himalaya. Environ. Earth Sci. 2019, 78, 231. [Google Scholar] [CrossRef]
- Lee, D.H.; Kim, Y.T.; Lee, S.R. Shallow landslide susceptibility models based on artificial neural networks considering the factor selection method and various non-linear activation functions. Remote Sens. 2020, 12, 1194. [Google Scholar] [CrossRef]
- He, L.; Coggan, J.; Francioni, M.; Eyre, M. Maximizing impacts of remote sensing surveys in slope stability—A novel method to incorporate discontinuities into machine learning landslide prediction. ISPRS Int. J. Geoinf. 2021, 10, 232. [Google Scholar] [CrossRef]
- Gao, Z.; Ding, M. Application of convolutional neural network fused with machine learning modeling framework for geospatial comparative analysis of landslide susceptibility. Nat. Hazards 2022, 113, 833–858. [Google Scholar] [CrossRef]
- Nguyen, P.T.; Tuyen, T.T.; Shirzadi, A.; Pham, B.T.; Shahabi, H.; Omidvar, E.; Amini, A.; Entezami, H.; Prakash, I.; Phong, T.V.; et al. Development of a novel hybrid intelligence approach for landslide spatial prediction. Appl. Sci. 2019, 9, 2824. [Google Scholar] [CrossRef]
- Novellino, A.; Cesarano, M.; Cappelletti, P.; Di Martire, D.; Di Napoli, M.; Ramondini, M.; Sowter, M.; Calcaterra, D. Slow-moving landslide risk assessment combining Machine Learning and InSAR techniques. Catena 2021, 203, 105317. [Google Scholar] [CrossRef]
- Youssef, A.M.; Pourghasemi, H.R.; Pourtaghi, Z.S.; Al-Katheeri, M.M. Landslide susceptibility mapping using random forest, boosted regression tree, classification and regression tree, and general linear models and comparison of their performance at Wadi Tayyah Basin, Asir Region, Saudi Arabia. Landslides 2016, 13, 839–856. [Google Scholar] [CrossRef]
- Kumar, D.; Thakur, M.; Dubey, C.S.; Shukla, D.P. Landslide susceptibility mapping & prediction using Support Vector Machine for Mandakini River Basin, Garhwal Himalaya, India. Geomorphology 2017, 295, 115–125. [Google Scholar]
- Roy, J.; Saha, S.; Arabameri, A.; Blaschke, T.; Bui, D.T. A novel ensemble approach for landslide susceptibility mapping (LSM) in Darjeeling and Kalimpong districts, West Bengal, India. Remote Sens. 2019, 11, 2866. [Google Scholar] [CrossRef]
- Ghorbanzadeh, O.; Blaschke, T.; Gholamnia, K.; Meena, S.R.; Tiede, D.; Aryal, J. Evaluation of different machine learning methods and deep-learning convolutional neural networks for landslide detection. Remote Sens. 2019, 11, 196. [Google Scholar] [CrossRef]
- Huang, F.; Zhang, J.; Zhou, C.; Wang, Y.; Huang, J.; Zhu, L. A deep learning algorithm using a fully connected sparse autoencoder neural network for landslide susceptibility prediction. Landslides 2020, 17, 217–229. [Google Scholar] [CrossRef]
- Liu, P.; Wei, Y.; Wang, Q.; Chen, Y.; Xie, J. Research on post-earthquake landslide extraction algorithm based on improved U-net model. Remote Sens. 2020, 12, 894. [Google Scholar] [CrossRef]
- Sharma, A.; Prakash, C.; Goshu, E.L.; Sharma, R. An artificial intelligence based framework to analyze the landside risk of a mountainous highway. Geocarto Int. 2023, 38, 2186494. [Google Scholar] [CrossRef]
- Alqadhi, S.; Mallick, J.; Alkahtani, M. Integrated deep learning with explainable artificial intelligence for enhanced landslide management. Nat. Hazards 2024, 120, 1343–1365. [Google Scholar] [CrossRef]
- Lee, S.; Ryu, J.H.; Min, K.; Won, J.S. Landslide susceptibility analysis using GIS and artificial neural network. Earth Surf. Process Landf. 2003, 28, 1361–1376. [Google Scholar] [CrossRef]
- Choi, J.; Oh, H.-J.; Lee, H.-J.; Lee, C.; Lee, S. Combining landslide susceptibility maps obtained from frequency ratio, logistic regression, and artificial neural network models using ASTER images and GIS. Eng. Geol. 2012, 124, 12–23. [Google Scholar] [CrossRef]
- Yuan, X.; Liu, C.; Nie, R.; Yang, Z.; Li, W.; Dai, X.; Cheng, J.; Zhang, J.; Ma, L.; Fu, X.; et al. A Comparative Analysis of Certainty Factor-Based Machine Learning Methods for Collapse and Landslide Susceptibility Mapping in Wenchuan County, China. Remote Sens. 2022, 14, 3259. [Google Scholar] [CrossRef]
- Dahim, M.; Alqadhi, S.; Mallick, J. Enhancing landslide management with hyper-tuned machine learning and deep learning models: Predicting susceptibility and analyzing sensitivity and uncertainty. Front. Ecol. Evol. 2023, 11, 1108924. [Google Scholar] [CrossRef]
- Sharma, N.; Saharia, M.; Ramana, G.V. High resolution landslide susceptibility mapping using ensemble machine learning and geospatial big data. Catena 2024, 235, 107653. [Google Scholar] [CrossRef]
- Song, Y.; Li, Y.; Zou, Y.; Wang, R.; Liang, Y.; Xu, S.; He, Y.; Yu, X.; Wu, W. Synergizing multiple machine learning techniques and remote sensing for advanced landslide susceptibility assessment: A case study in the Three Gorges Reservoir Area. Environ. Earth Sci. 2024, 83, 227. [Google Scholar] [CrossRef]
- Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Moher, D. Updating guidance for reporting systematic reviews: Development of the PRISMA 2020 statement. J. Clin. Epidemiol. 2021, 134, 103–112. [Google Scholar] [CrossRef]
- Han, J.; Kang, H.J.; Kim, M.; Kwon, G.H. Mapping the intellectual structure of research on surgery with mixed reality: Bibliometric network analysis (2000–2019). J. Biomed. Inform. 2020, 109, 103516. [Google Scholar] [CrossRef]
- Van Eck, N.J.; Waltman, L. Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics 2010, 84, 523–538. [Google Scholar] [CrossRef]
- Carlberg, C.; Carlberg, C.G. Statistical Analysis: Microsoft Excel 2013; Que Publishing: Indianapolis, IN, USA, 2014. [Google Scholar]
- Morante-Carballo, F.; Montalván-Burbano, N.; Carrión-Mero, P.; Jácome-Francis, K. Worldwide Research Analysis on Natural Zeolites as Environmental Remediation Materials. Sustainability 2021, 13, 6378. [Google Scholar] [CrossRef]
- Sun, H.; Li, W.; Scaioni, M.; Fu, J.; Guo, X.; Gao, J. Influence of spatial heterogeneity on landslide susceptibility in the transboundary area of the Himalayas. Geomorphology 2023, 433, 108723. [Google Scholar] [CrossRef]
- Arabameri, A.; Saha, S.; Roy, J.; Chen, W.; Blaschke, T.; Bui, D.T. Landslide susceptibility evaluation and management using different machine learning methods in the Gallicash River Watershed, Iran. Remote Sens. 2020, 12, 475. [Google Scholar] [CrossRef]
- Yang, K.; Li, W.; Yang, X.; Zhang, L. Improving Landslide Recognition on UAV Data through Transfer Learning. Appl. Sci. 2022, 12, 10121. [Google Scholar] [CrossRef]
- Masruroh, H.; Leksono, A.S.; Kurniawan, S. Developing landslide susceptibility map using Artificial Neural Network (ANN) method for mitigation of land degradation. J. Degrad. Min. Lands Manag. 2023, 10, 4479–4494. [Google Scholar] [CrossRef]
- Sun, D.; Ding, Y.; Zhang, J.; Wen, H.; Wang, Y.; Xu, J.; Zhou, X.; Liu, R. Essential insights into decision mechanism of landslide susceptibility mapping based on different machine learning models. Geocarto Int. 2022, 38, 1–29. [Google Scholar] [CrossRef]
- Das, R.; Wegmann, K.W. Evaluation of machine learning-based algorithms for landslide detection across satellite sensors for the 2019 Cyclone Idai event, Chimanimani District, Zimbabwe. Landslides 2022, 19, 2965–2981. [Google Scholar] [CrossRef]
- Hu, Q.; Zhou, Y.; Wang, S.; Wang, F.; Wang, H. Improving the accuracy of landslide detection in “off-site” area by machine learning model portability comparison: A case study of Jiuzhaigou Earthquake, China. Remote Sens. 2019, 11, 2530. [Google Scholar] [CrossRef]
- Xie, W.; Li, X.; Jian, W.; Yang, Y.; Liu, H.; Robledo, L.F.; Nie, W. A novel hybrid method for landslide susceptibility mapping-based geodetector and machine learning cluster: A case of Xiaojin County, China. ISPRS Int. J. Geoinf. 2021, 10, 93. [Google Scholar] [CrossRef]
- Nguyen, H.D.; Nguyen, Q.H.; Du, Q.V.V.; Pham, V.T.; Pham, L.T.; Van Hoang, T.; Truong, Q.H.; Bui, Q.T.; Petrisor, A.I. Landslide susceptibility prediction using machine learning and remote sensing: Case study in Thua Thien Hue province, Vietnam. Geol. J. 2024, 59, 636–658. [Google Scholar] [CrossRef]
- Ma, Y.; Xu, S.; Jiang, T.; Wang, Z.; Wang, Y.; Liu, M.; Li, X.; Ma, X. Learning a Deep Attention Dilated Residual Convolutional Neural Network for Landslide Susceptibility Mapping in Hanzhong City, Shaanxi Province, China. Remote Sens. 2023, 15, 3296. [Google Scholar] [CrossRef]
- Sheng, Y.; Xu, G.; Jin, B.; Zhou, C.; Li, Y.; Chen, W. Data-Driven Landslide Spatial Prediction and Deformation Monitoring: A Case Study of Shiyan City, China. Remote Sens. 2023, 15, 5256. [Google Scholar] [CrossRef]
- Ji, S.; Yu, D.; Shen, C.; Li, W.; Xu, Q. Landslide detection from an open satellite imagery and digital elevation model dataset using attention boosted convolutional neural networks. Landslides 2020, 17, 1337–1352. [Google Scholar] [CrossRef]
- Xiong, Y.; Zhou, Y.; Wang, F.; Wang, S.; Wang, J.; Ji, J.; Wang, Z. Landslide Susceptibility Mapping Using Ant Colony Optimization Strategy and Deep Belief Network in Jiuzhaigou Region. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2021, 14, 11042–11057. [Google Scholar] [CrossRef]
- Habumugisha, J.M.; Chen, N.; Rahman, M.; Islam, M.M.; Ahmad, H.; Elbeltagi, A.; Sharma, G.; Liza, S.N.; Dewan, A. Landslide Susceptibility Mapping with Deep Learning Algorithms. Sustainability 2022, 14, 1734. [Google Scholar] [CrossRef]
- Wang, J.; Jaboyedoff, M.; Chen, G.; Luo, X.; Derron, M.H.; Hu, Q.; Fei, L.; Prajapati, G.; Choanji, T.; Luo, S.; et al. Landslide susceptibility prediction and mapping using the LD-BiLSTM model in seismically active mountainous regions. Landslides 2024, 21, 17–34. [Google Scholar] [CrossRef]
- Niu, R.; Wu, X.; Yao, X.; Peng, L.; Ai, L.; Peng, J. Susceptibility assessment of landslides triggered by the Lushan earthquake, April 20, 2013, China. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2014, 7, 3979–3992. [Google Scholar] [CrossRef]
- Tien Bui, D.; Tuan, T.A.; Hoang, N.D.; Thanh, N.Q.; Nguyen, D.B.; Van Liem, N.; Pradhan, B. Spatial prediction of rainfall-induced landslides for the Lao Cai area (Vietnam) using a hybrid intelligent approach of least squares support vector machines inference model and artificial bee colony optimization. Landslides 2017, 14, 447–458. [Google Scholar] [CrossRef]
- Pourghasemi, H.R.; Gayen, A.; Park, S.; Lee, C.W.; Lee, S. Assessment of landslide-prone areas and their zonation using logistic regression, LogitBoost, and naïvebayes machine-learning algorithms. Sustainability 2018, 10, 3697. [Google Scholar] [CrossRef]
- Arabameri, A.; Pradhan, B.; Rezaei, K.; Sohrabi, M.; Kalantari, Z. GIS-based landslide susceptibility mapping using numerical risk factor bivariate model and its ensemble with linear multivariate regression and boosted regression tree algorithms. J. Mt. Sci. 2019, 16, 595–618. [Google Scholar] [CrossRef]
- Kocaman, S.; Tavus, B.; Nefeslioglu, H.A.; Karakas, G.; Gokceoglu, C. Evaluation of Floods and Landslides Triggered by a Meteorological Catastrophe (Ordu, Turkey, August 2018) Using Optical and Radar Data. Geofluids 2020, 1, 8830661. [Google Scholar] [CrossRef]
- Al-Najjar, H.A.H.; Pradhan, B.; Sarkar, R.; Beydoun, G.; Alamri, A. A new integrated approach for landslide data balancing and spatial prediction based on generative adversarial networks (GAN). Remote Sens. 2021, 13, 4011. [Google Scholar] [CrossRef]
- Khalil, Y.M.; Al-Masnay, Y.A.; Al-Areeq, N.M.; Al-Aizari, A.R.; Al-Shaibah, B.; Liu, X. Estimating landslide hazard distribution based on machine learning and bivariate statistics in Utmah Region, Yemen. Nat. Hazards 2024, 120, 2869–2907. [Google Scholar] [CrossRef]
- Huang, F.; Tao, S.; Li, D.; Lian, Z.; Catani, F.; Huang, J.; Li, K.; Zhang, C. Landslide Susceptibility Prediction Considering Neighborhood Characteristics of Landslide Spatial Datasets and Hydrological Slope Units Using Remote Sensing and GIS Technologies. Remote Sens. 2022, 14, 4436. [Google Scholar] [CrossRef]
- Hussain, M.A.; Chen, Z.; Kalsoom, I.; Asghar, A.; Shoaib, M. Landslide Susceptibility Mapping Using Machine Learning Algorithm: A Case Study Along Karakoram Highway (KKH), Pakistan. J. Indian Soc. Remote Sens. 2022, 50, 849–866. [Google Scholar] [CrossRef]
- Chowdhuri, I.; Pal, S.C.; Saha, A.; Chakrabortty, R.; Roy, P. Mapping of earthquake hotspot and coldspot zones for identifying potential landslide hotspot areas in the Himalayan region. Bull. Eng. Geol. Environ. 2022, 81, 257. [Google Scholar] [CrossRef]
- Zhang, T.; Zhang, W.; Cao, D.; Yi, Y.; Wu, X. A New Deep Learning Neural Network Model for the Identification of InSAR Anomalous Deformation Areas. Remote Sens. 2022, 14, 2690. [Google Scholar] [CrossRef]
- Deng, N.; Li, Y.; Ma, J.; Shahabi, H.; Hashim, M.; de Oliveira, G.; Chaeikar, S.S. A comparative study for landslide susceptibility assessment using machine learning algorithms based on grid unit and slope unit. Front. Environ. Sci. 2022, 10, 1009433. [Google Scholar] [CrossRef]
- Zeng, B.; Chen, X. Assessment of shallow landslide susceptibility using an artificial neural network. Arab. J. Geosci. 2021, 14, 499. [Google Scholar] [CrossRef]
- Fang, H.; Shao, Y.; Xie, C.; Tian, B.; Shen, C.; Zhu, Y.; Guo, Y.; Yang, Y.; Chen, G.; Zhang, M. A New Approach to Spatial Landslide Susceptibility Prediction in Karst Mining Areas Based on Explainable Artificial Intelligence. Sustainability 2023, 15, 3094. [Google Scholar] [CrossRef]
- Zheng, X.; Han, L.; He, G.; Wang, N.; Wang, G.; Feng, L. Semantic Segmentation Model for Wide-Area Coseismic Landslide Extraction Based on Embedded Multichannel Spectral–Topographic Feature Fusion: A Case Study of the Jiuzhaigou Ms7.0 Earthquake in Sichuan, China. Remote Sens. 2023, 15, 1084. [Google Scholar] [CrossRef]
- Ghorbanzadeh, O.; Meena, S.R.; Blaschke, T.; Aryal, J. UAV-based slope failure detection using deep-learning convolutional neural networks. Remote Sens. 2019, 11, 2046. [Google Scholar] [CrossRef]
- Chen, X.; Zhao, C.; Xi, J.; Lu, Z.; Ji, S.; Chen, L. Deep Learning Method of Landslide Inventory Map with Imbalanced Samples in Optical Remote Sensing. Remote Sens. 2022, 14, 5517. [Google Scholar] [CrossRef]
- Das, S.; Sharma, P.; Pain, A.; Kanungo, D.P.; Sarkar, S. Deep learning based landslide detection using open-source resources: Opportunities and challenges. Earth Sci. Inform. 2023, 16, 4035–4052. [Google Scholar] [CrossRef]
- Yin, G.; Luo, J.; Niu, F.; Lin, Z.; Liu, M. Machine learning-based thermokarst landslide susceptibility modeling across the permafrost region on the Qinghai-Tibet Plateau. Landslides 2021, 18, 2639–2649. [Google Scholar] [CrossRef]
- Qasimi, A.B.; Isazade, V.; Enayat, E.; Nadry, Z.; Majidi, A.H. Landslide susceptibility mapping in Badakhshan province, Afghanistan: A comparative study of machine learning algorithms. Geocarto Int. 2023, 38, 2248082. [Google Scholar] [CrossRef]
- Agboola, G.; Beni, L.H.; Elbayoumi, T.; Thompson, G. Optimizing landslide susceptibility mapping using machine learning and geospatial techniques. Ecol. Inform. 2024, 81, 102583. [Google Scholar] [CrossRef]
- Patil, A.S.; Panhalkar, S.S. A holistic approach of remote sensing, GIS, and machine learning for shallow landslide susceptibility mapping in Gaganbawada region of Western Ghats, India. Proc. Indian Natl. Sci. Acad. 2024, 1–18. [Google Scholar] [CrossRef]
- Stumpf, A.; Kerle, N. Object-oriented mapping of landslides using Random Forests. Remote Sens. Environ. 2011, 115, 2564–2577. [Google Scholar] [CrossRef]
- Dou, J.; Yunus, A.P.; Bui, D.T.; Merghadi, A.; Sahana, M.; Zhu, Z.; Chen, C.W.; Han, Z.; Pham, B.T. Improved landslide assessment using support vector machine with bagging, boosting, and stacking ensemble machine learning framework in a mountainous watershed, Japan. Landslides 2020, 17, 641–658. [Google Scholar] [CrossRef]
- Shahabi, H.; Rahimzad, M.; Piralilou, S.T.; Ghorbanzadeh, O.; Homayouni, S.; Blaschke, T.; Lim, S.; Ghamisi, P. Unsupervised deep learning for landslide detection from multispectral sentinel-2 imagery. Remote Sens. 2021, 13, 4698. [Google Scholar] [CrossRef]
- Li, H.; He, Y.; Xu, Q.; Deng, J.; Li, W.; Wei, Y. Detection and segmentation of loess landslides via satellite images: A two-phase framework. Landslides 2022, 19, 673–686. [Google Scholar] [CrossRef]
- Sweta, K.; Goswami, A.; Peethambaran, B.; Bahuguna, I.M.; Rajawat, A.S. Landslide susceptibility zonation around Dharamshala, Himachal Pradesh, India: An artificial intelligence model–based assessment. Bull. Eng. Geol. Environ. 2022, 81, 310. [Google Scholar] [CrossRef]
- Zhang, Y.; Deng, L.; Han, Y.; Sun, Y.; Zang, Y.; Zhou, M. Landslide Hazard Assessment in Highway Areas of Guangxi Using Remote Sensing Data and a Pre-Trained XGBoost Model. Remote Sens. 2023, 15, 3350. [Google Scholar] [CrossRef]
- Nath, S.K.; Sengupta, A.; Srivastava, A. Remote sensing GIS-based landslide susceptibility & risk modeling in Darjeeling–Sikkim Himalaya together with FEM-based slope stability analysis of the terrain. Nat. Hazards 2021, 108, 3271–3304. [Google Scholar]
- Garcia, G.P.B.; Soares, L.P.; Espadoto, M.; Grohmann, C.H. Relict landslide detection using deep-learning architectures for image segmentation in rainforest areas: A new framework. Int. J. Remote Sens. 2023, 44, 2168–2195. [Google Scholar] [CrossRef]
- Pourghasemi, H.R.; Goli Jirandeh, A.; Pradhan, B.; Xu, C.; Gokceoglu, C. Landslide susceptibility mapping using support vector machine and GIS at the Golestan Province, Iran. J. Earth Sys. Sci. 2013, 122, 349–369. [Google Scholar] [CrossRef]
- Qing, F.; Zhao, Y.; Meng, X.; Su, X.; Qi, T.; Yue, D. Application of machine learning to debris flow susceptibility mapping along the China-Pakistan Karakoram Highway. Remote Sens. 2020, 12, 2933. [Google Scholar] [CrossRef]
- Ahmad, H.; Ningsheng, C.; Rahman, M.; Islam, M.M.; Pourghasemi, H.R.; Hussain, S.F.; Habumugisha, J.M.; Liu, E.; Zheng, H.; Ni, H.; et al. Geohazards susceptibility assessment along the upper indus basin using four machine learning and statistical models. ISPRS Int. J. Geoinf. 2021, 10, 315. [Google Scholar] [CrossRef]
- Al-Najjar, H.A.H.; Pradhan, B.; Kalantar, B.; Sameen, M.I.; Santosh, M.; Alamri, A. Landslide susceptibility modeling: An integrated novel method based on machine learning feature transformation. Remote Sens. 2021, 13, 3281. [Google Scholar] [CrossRef]
- Chang, L.; Zhang, R.; Wang, C. Evaluation and Prediction of Landslide Susceptibility in Yichang Section of Yangtze River Basin Based on Integrated Deep Learning Algorithm. Remote Sens. 2022, 14, 2717. [Google Scholar] [CrossRef]
- Huang, C.; Li, F.; Wei, L.; Hu, X.; Yang, Y. Landslide Susceptibility Modeling Using a Deep Random Neural Network. Appl. Sci. 2022, 12, 12887. [Google Scholar] [CrossRef]
- Shahzad, N.; Ding, X.; Abbas, S. A Comparative Assessment of Machine Learning Models for Landslide Susceptibility Mapping in the Rugged Terrain of Northern Pakistan. Appl. Sci. 2022, 12, 2280. [Google Scholar] [CrossRef]
- Wang, X.; Zhang, X.; Bi, J.; Zhang, X.; Deng, S.; Liu, Z.; Wang, L.; Guo, H. Landslide Susceptibility Evaluation Based on Potential Disaster Identification and Ensemble Learning. Int. J. Environ. Res. Public Health 2022, 19, 14241. [Google Scholar] [CrossRef]
- Zhang, H.; Song, Y.; Xu, S.; He, Y.; Li, Z.; Yu, X.; Liang, Y.; Wu, W.; Wang, Y. Combining a class-weighted algorithm and machine learning models in landslide susceptibility mapping: A case study of Wanzhou section of the Three Gorges Reservoir, China. Comput. Geosci. 2022, 158, 104966. [Google Scholar] [CrossRef]
- Zhang, S.; Bai, L.; Li, Y.; Li, W.; Xie, M. Comparing Convolutional Neural Network and Machine Learning Models in Landslide Susceptibility Mapping: A Case Study in Wenchuan County. Front. Environ. Sci. 2022, 10, 886841. [Google Scholar] [CrossRef]
- Cao, W.G.; Fu, Y.; Dong, Q.Y.; Wang, H.G.; Ren, Y.; Li, Z.Y.; Du, Y.Y. Landslide susceptibility assessment in Western Henan Province based on a comparison of conventional and ensemble machine learning. China Geol. 2023, 6, 409–419. [Google Scholar]
- Chang, Z.; Catani, F.; Huang, F.; Liu, G.; Meena, S.R.; Huang, J.; Zhou, C. Landslide susceptibility prediction using slope unit-based machine learning models considering the heterogeneity of conditioning factors. J. Rock Mech. Geotech. Eng. 2023, 15, 1127–1143. [Google Scholar] [CrossRef]
- Miao, F.; Zhao, F.; Wu, Y.; Li, L.; Török, Á. Landslide susceptibility mapping in Three Gorges Reservoir area based on GIS and boosting decision tree model. Stoch. Environ. Res. Risk Assess. 2023, 37, 2283–2303. [Google Scholar] [CrossRef]
- Parra, F.; González, J.; Chacón, M.; Marín, M. Modeling and Evaluation of the Susceptibility to Landslide Events Using Machine Learning Algorithms in the Province of Chañaral, Atacama Region, Chile. Sustainability 2023, 15, 16806. [Google Scholar] [CrossRef]
- Wang, S.; Ling, S.; Wu, X.; Wen, H.; Huang, J.; Wang, F.; Sun, C. Key predisposing factors and susceptibility assessment of landslides along the Yunnan–Tibet traffic corridor, Tibetan plateau: Comparison with the LR, RF, NB, and MLP techniques. Front. Earth Sci. 2023, 10, 1100363. [Google Scholar] [CrossRef]
- Abbas, F.; Zhang, F.; Hussain, M.A.; Abbas, H.; Alrefaei, A.F.; Albeshr, M.F.; Iqbal, J.; Ghani, J.; Ismail, S. Landslide susceptibility assessment along the Karakoram highway, Gilgit Baltistan, Pakistan: A comparative study between ensemble and neighbor-based machine learning algorithms. Sci. Remote Sens. 2024, 9, 100132. [Google Scholar] [CrossRef]
- Chen, Z.; Quan, H.; Jin, R.; Jin, A.; Lin, Z.; Jin, G.; Jin, G. Assessment of Landslide Susceptibility Using the PCA and ANFIS with Various Metaheuristic Algorithms. KSCE J. Civ. Eng. 2024, 28, 1461–1474. [Google Scholar] [CrossRef]
- Huang, F.; Li, R.; Catani, F.; Zhou, X.; Zeng, Z.; Huang, J. Uncertainties in landslide susceptibility prediction: Influence rule of different levels of errors in landslide spatial position. J. Rock Mech.Geotech. Eng. 2024, in press. [Google Scholar] [CrossRef]
- Huang, F.; Teng, Z.; Yao, C.; Jiang, S.H.; Catani, F.; Chen, W.; Huang, J. Uncertainties of landslide susceptibility prediction: Influences of random errors in landslide conditioning factors and errors reduction by low pass filter method. J. Rock Mech. Geotech. Eng. 2024, 16, 213–230. [Google Scholar] [CrossRef]
- Lu, D.; Tang, G.; Yan, G.; Yu, F.; Lin, X. Comparison of different open-source Digital Elevation Models for landslide susceptibility mapping. Earth Surf. Process. Landf. 2024, 49, 1411–1427. [Google Scholar] [CrossRef]
- Sun, D.; Wang, J.; Wen, H.; Ding, Y.; Gu, Q.; Zhang, J.; Zhang, F. Insights into landslide susceptibility in different karst erosion landforms based on interpretable machine learning. Earth Surf. Process. Landf. 2024, 49, 1006–1027. [Google Scholar] [CrossRef]
- Zeng, T.; Wu, L.; Hayakawa, Y.S.; Yin, K.; Gui, L.; Jin, B.; Guo, Z.; Peduto, D. Advanced integration of ensemble learning and MT-InSAR for enhanced slow-moving landslide susceptibility zoning. Eng. Geol. 2024, 331, 107436. [Google Scholar] [CrossRef]
- Zhao, S.; Zeng, R.; Zhang, Z.; Meng, X.; Qi, T.; Long, Z.; Guo, W.; Si, G. Landslide Mapping and Causes of Landslides in the China–Nepal Transportation Corridor Based on Remote Sensing Technology. Remote Sens. 2024, 16, 356. [Google Scholar] [CrossRef]
- Zhou, C.; Cao, Y.; Hu, X.; Yin, K.; Wang, Y.; Catani, F. Enhanced dynamic landslide hazard mapping using MT-InSAR method in the Three Gorges Reservoir Area. Landslides 2022, 19, 1585–1597. [Google Scholar] [CrossRef]
- Xiao, L.; Zhang, Y.; Peng, G. Landslide susceptibility assessment using integrated deep learning algorithm along the China-Nepal highway. Sensors 2018, 18, 4436. [Google Scholar] [CrossRef]
- Pham, B.T.; Jaafari, A.; Prakash, I.; Bui, D.T. A novel hybrid intelligent model of support vector machines and the MultiBoost ensemble for landslide susceptibility modeling. Bull. Eng. Geol. Environ. 2019, 78, 2865–2886. [Google Scholar] [CrossRef]
- Hu, Q.; Zhou, Y.; Wang, S.; Wang, F. Machine learning and fractal theory models for landslide susceptibility mapping: Case study from the Jinsha River Basin. Geomorphology 2020, 351, 106975. [Google Scholar] [CrossRef]
- Huang, F.; Cao, Z.; Guo, J.; Jiang, S.H.; Li, S.; Guo, Z. Comparisons of heuristic, general statistical and machine learning models for landslide susceptibility prediction and mapping. Catena 2020, 191, 104580. [Google Scholar] [CrossRef]
- Huang, F.; Cao, Z.; Jiang, S.H.; Zhou, C.; Huang, J.; Guo, Z. Landslide susceptibility prediction based on a semi-supervised multiple-layer perceptron model. Landslides 2020, 17, 2919–2930. [Google Scholar] [CrossRef]
- Abu El-Magd, S.A.; Ali, S.A.; Pham, Q.B. Spatial modeling and susceptibility zonation of landslides using random forest, naïve bayes and K-nearest neighbor in a complicated terrain. Earth Sci. Inform. 2021, 14, 1227–1243. [Google Scholar] [CrossRef]
- Bruzón, A.G.; Arrogante-Funes, P.; Arrogante-Funes, F.; Martín-González, F.; Novillo, C.J.; Fernández, R.R.; Vázquez-Jiménez, R.; Alarcón-Paredes, A.; Alonso-Silverio, G.A.; Cantu-Ramirez, C.A.; et al. Landslide susceptibility assessment using an automl framework. Int. J. Environ. Res. Public Health 2021, 18, 10971. [Google Scholar] [CrossRef]
- Kong, C.; Tian, Y.; Ma, X.; Weng, Z.; Zhang, Z.; Xu, K. Landslide susceptibility assessment based on different machine learning methods in Zhaoping county of eastern Guangxi. Remote Sens. 2021, 13, 3573. [Google Scholar] [CrossRef]
- Lin, Y.T.; Chen, Y.K.; Yang, K.H.; Chen, C.S.; Han, J.Y. Integrating InSAR observables and multiple geological factors for landslide susceptibility assessment. Appl. Sci. 2021, 11, 7289. [Google Scholar] [CrossRef]
- Huang, F.; Ye, Z.; Zhou, X.; Huang, J.; Zhou, C. Landslide susceptibility prediction using an incremental learning Bayesian Network model considering the continuously updated landslide inventories. Bull. Eng. Geol. Environ. 2022, 81, 250. [Google Scholar] [CrossRef]
- Mabdeh, A.N.; Al-Fugara, A.K.; Ahmadlou, M.; Al-Adamat, R.; Al-Shabeeb, A.R. GIS-based landslide susceptibility assessment and mapping in Ajloun and Jerash governorates in Jordan using genetic algorithm-based ensemble models. Acta Geophys. 2022, 70, 1253–1267. [Google Scholar] [CrossRef]
- Peng, T.; Chen, Y.; Chen, W. Landslide Susceptibility Modeling Using Remote Sensing Data and Random SubSpace-Based Functional Tree Classifier. Remote Sens. 2022, 14, 4803. [Google Scholar] [CrossRef]
- Ullah, I.; Aslam, B.; Shah, S.H.I.A.; Tariq, A.; Qin, S.; Majeed, M.; Havenith, H.B. An Integrated Approach of Machine Learning, Remote Sensing, and GIS Data for the Landslide Susceptibility Mapping. Land 2022, 11, 1265. [Google Scholar] [CrossRef]
- Wei, R.; Ye, C.; Sui, T.; Ge, Y.; Li, Y.; Li, J. Combining spatial response features and machine learning classifiers for landslide susceptibility mapping. Int. J. Appl. Earth Obs. Geoinf. 2022, 107, 102681. [Google Scholar] [CrossRef]
- Asghar, A.; Su, L.J.; Zhao, B.; Usmani, N.A. Integrating predictive modeling techniques with geospatial data for landslide susceptibility assessment in northern Pakistan. J. Mt. Sci. 2023, 20, 2603–2627. [Google Scholar] [CrossRef]
- Hussain, M.A.; Chen, Z.; Zheng, Y.; Zhou, Y.; Daud, H. Deep Learning and Machine Learning Models for Landslide Susceptibility Mapping with Remote Sensing Data. Remote Sens. 2023, 15, 4703. [Google Scholar] [CrossRef]
- Shahabi, H.; Ahmadi, R.; Alizadeh, M.; Hashim, M.; Al-Ansari, N.; Shirzadi, A.; Wolf, I.D.; Ariffin, E.H. Landslide Susceptibility Mapping in a Mountainous Area Using Machine Learning Algorithms. Remote Sens. 2023, 15, 3122. [Google Scholar] [CrossRef]
- Wang, X.; Du, A.; Hu, F.; Liu, Z.; Zhang, X.; Wang, L.; Guo, H. Landslide susceptibility evaluation based on active deformation and graph convolutional network algorithm. Front. Earth Sci. 2023, 11, 1132722. [Google Scholar] [CrossRef]
- Wang, Z.; Xu, S.; Liu, J.; Wang, Y.; Ma, X.; Jiang, T.; He, X.; Han, Z. A Combination of Deep Autoencoder and Multi-Scale Residual Network for Landslide Susceptibility Evaluation. Remote Sens. 2023, 15, 653. [Google Scholar] [CrossRef]
- Zeng, H.; Zhu, Q.; Ding, Y.; Hu, H.; Chen, L.; Xie, X.; Chen, M.; Yao, Y. Graph neural networks with constraints of environmental consistency for landslide susceptibility evaluation. Int. J. Geogr. Inf. Sci. 2022, 36, 2270–2295. [Google Scholar] [CrossRef]
- Chen, Y.; Ming, D.; Yu, J.; Xu, L.; Ma, Y.; Li, Y.; Ling, X.; Zhu, Y. Susceptibility-Guided Landslide Detection Using Fully Convolutional Neural Network. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2023, 16, 998–1018. [Google Scholar] [CrossRef]
- Zhang, X.; Yu, W.; Pun, M.O.; Shi, W. Cross-domain landslide mapping from large-scale remote sensing images using prototype-guided domain-aware progressive representation learning. ISPRS J. Photogramm. Remote Sens. 2023, 197, 1–17. [Google Scholar] [CrossRef]
- Ouyang, S.; Chen, W.; Liu, H.; Li, Y.; Xu, Z. A novel landslide susceptibility prediction framework based on contrastive loss. GISci. Remote Sens. 2024, 61, 2306740. [Google Scholar] [CrossRef]
- Sameen, M.I.; Pradhan, B. Landslide Detection Using Residual Networks and the Fusion of Spectral and Topographic Information. IEEE Access 2019, 7, 114363–114373. [Google Scholar] [CrossRef]
- Chen, T.H.K.; Kincey, M.E.; Rosser, N.J.; Seto, K.C. Identifying recurrent and persistent landslides using satellite imagery and deep learning: A 30-year analysis of the Himalaya. Sci. Total Environ. 2024, 922, 171161. [Google Scholar] [CrossRef] [PubMed]
- Cheng, L.; Li, J.; Duan, P.; Wang, M. A small attentional YOLO model for landslide detection from satellite remote sensing images. Landslides 2021, 18, 2751–2765. [Google Scholar] [CrossRef]
- Shi, W.; Zhang, M.; Ke, H.; Fang, X.; Zhan, Z.; Chen, S. Landslide Recognition by Deep Convolutional Neural Network and Change Detection. IEEE Trans. Geosci. Remote Sens. 2021, 59, 4654–4672. [Google Scholar] [CrossRef]
- Singh, P.; Maurya, V.; Dwivedi, R. Pixel-based landslide identification using Landsat 8 and GEE. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci.-ISPRS Arch. 2021, 43, 721–726. [Google Scholar] [CrossRef]
- Ullo, S.; Mohan, A.; Sebastianelli, A.; Ahamed, S.; Kumar, B.; Dwivedi, R.; Sinha, G. A New Mask R-CNN-Based Method for Improved Landslide Detection. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2021, 14, 3799–3810. [Google Scholar] [CrossRef]
- Jiang, W.; Xi, J.; Li, Z.; Zang, M.; Chen, B.; Zhang, C.; Liu, Z.; Gao, S.; Zhu, W. Deep Learning for Landslide Detection and Segmentation in High-Resolution Optical Images along the Sichuan-Tibet Transportation Corridor. Remote Sens. 2022, 14, 5490. [Google Scholar] [CrossRef]
- Ghorbanzadeh, O.; Xu, Y.; Ghamisi, P.; Kopp, M.; Kreil, D. Landslide4Sense: Reference Benchmark Data and Deep Learning Models for Landslide Detection. IEEE Trans. Geosci. Remote Sens. 2022, 60, 5633017. [Google Scholar] [CrossRef]
- Meena, S.R.; Soares, L.P.; Grohmann, C.H.; Van Westen, C.; Bhuyan, K.; Singh, R.P.; Floris, M.; Catani, F. Landslide detection in the Himalayas using machine learning algorithms and U-Net. Landslides 2022, 19, 1209–1229. [Google Scholar] [CrossRef]
- Tang, X.; Tu, Z.; Wang, Y.; Liu, M.; Li, D.; Fan, X. Automatic Detection of Coseismic Landslides Using a New Transformer Method. Remote Sens. 2022, 14, 2884. [Google Scholar] [CrossRef]
- Yang, S.; Wang, Y.; Wang, P.; Mu, J.; Jiao, S.; Zhao, X.; Wang, Z.; Wang, K.; Zhu, Y. Automatic Identification of Landslides Based on Deep Learning. Appl. Sci. 2022, 12, 8153. [Google Scholar] [CrossRef]
- Yu, B.; Xu, C.; Chen, F.; Wang, N.; Wang, L. HADeenNet: A hierarchical-attention multi-scale deconvolution network for landslide detection. Int. J. Appl. Earth Obs. Geoinf. 2022, 111, 102853. [Google Scholar] [CrossRef]
- Chandra, N.; Sawant, S.; Vaidya, H. An Efficient U-Net Model for Improved Landslide Detection from Satellite Images. PFG—J. Photogramm. Remote Sens. Geoinf. Sci. 2023, 91, 13–28. [Google Scholar] [CrossRef]
- Guan, Y.; Yu, L.; Hao, S.; Li, L.; Zhang, X.; Hao, M. Slope Failure and Landslide Detection in Huangdao District of Qingdao City Based on an Improved Faster R-CNN Model. GeoHazards 2023, 4, 302–315. [Google Scholar] [CrossRef]
- Li, D.; Tang, X.; Tu, Z.; Fang, C.; Ju, Y. Automatic Detection of Forested Landslides: A Case Study in Jiuzhaigou County, China. Remote Sens. 2023, 15, 3850. [Google Scholar] [CrossRef]
- Lv, P.; Ma, L.; Li, Q.; Du, F. ShapeFormer: A Shape-Enhanced Vision Transformer Model for Optical Remote Sensing Image Landslide Detection. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2023, 16, 2681–2689. [Google Scholar] [CrossRef]
- Wang, J.; Chen, G.; Jaboyedoff, M.; Derron, M.H.; Li, F.; Li, H.; Luo, X. Loess landslides detection via a partially supervised learning and improved Mask-RCNN with multi-source remote sensing data. Catena 2023, 231, 107371. [Google Scholar] [CrossRef]
- Yang, Q.; Wang, X.; Zhang, X.; Zheng, J.; Ke, Y.; Wang, L.; Guo, H. A Novel Deep Learning Method for Automatic Recognition of Coseismic Landslides. Remote Sens. 2023, 15, 977. [Google Scholar] [CrossRef]
- Chen, H.; He, Y.; Zhang, L.; Yang, W.; Liu, Y.; Gao, B.; Zhang, Q.; Lu, J. A Multi-Input Channel U-Net Landslide Detection Method Fusing SAR Multisource Remote Sensing Data. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17, 1215–1232. [Google Scholar] [CrossRef]
- Dang, K.B.; Nguyen, C.Q.; Tran, Q.C.; Nguyen, H.; Nguyen, T.T.; Nguyen, D.A.; Tran, T.H.; Bui, P.T.; Giang, T.L.; Lenh, T.A.; et al. Comparison between U-shaped structural deep learning models to detect landslide traces. Sci. Total Environ. 2024, 912, 169113. [Google Scholar] [CrossRef]
- Cai, H.; Chen, T.; Niu, R.; Plaza, A. Landslide Detection Using Densely Connected Convolutional Networks and Environmental Conditions. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2021, 14, 5235–5247. [Google Scholar]
- Huang, R.; Chen, T. Landslide Recognition from Multi-Feature Remote Sensing Data Based on Improved Transformers. Remote Sens. 2023, 15, 3340. [Google Scholar] [CrossRef]
- Sreelakshmi, S.; Vinod Chandra, S.S. Visual saliency-based landslide identification using super-resolution remote sensing data. Results Eng. 2024, 21, 101656. [Google Scholar] [CrossRef]
- Van Natijne, A.L.; Bogaard, T.A.; Zieher, T.; Pfeiffer, J.; Lindenbergh, R.C. Machine-learning-based nowcasting of the Vögelsberg deep-seated landslide: Why predicting slow deformation is not so easy. Nat. Hazards Earth Sys. Sci. 2023, 23, 3723–3745. [Google Scholar] [CrossRef]
- Chen, L.; Peng, L.; Yang, L. Improving Landslide Prediction: Innovative Modeling and Evaluation of Landslide Scenario with Knowledge Graph Embedding. Remote Sens. 2024, 16, 145. [Google Scholar] [CrossRef]
- Sachdeva, S.; Bhatia, T.; Verma, A.K. A novel voting ensemble model for spatial prediction of landslides using GIS. Int. J. Remote Sens. 2020, 41, 929–952. [Google Scholar] [CrossRef]
- Chen, L.; Ge, X.; Yang, L.; Li, W.; Peng, L. An Improved Multi-Source Data-Driven Landslide Prediction Method Based on Spatio-Temporal Knowledge Graph. Remote Sens. 2023, 15, 2126. [Google Scholar] [CrossRef]
- Sundriyal, Y.; Kumar, S.; Chauhan, N.; Kaushik, S.; Kumar, V.; Rana, N.; Wasson, R. An integrated approach of machine learning and remote sensing for evaluating landslide hazards and risk hotspots, NW Himalaya. Remote Sens. Appl. Soci. Environ. 2024, 33, 101140. [Google Scholar] [CrossRef]
- Zhou, C.; Cao, Y.; Gan, L.; Wang, Y.; Motagh, M.; Roessner, S.; Hu, X.; Yin, K. A novel framework for landslide displacement prediction using MT-InSAR and machine learning techniques. Eng. Geol. 2024, 334, 107497. [Google Scholar] [CrossRef]
- Maxwell, A.E.; Sharma, M.; Donaldson, K.A. Explainable boosting machines for slope failure spatial predictive modeling. Remote Sens. 2021, 13, 4991. [Google Scholar] [CrossRef]
- Huang, F.; Xiong, H.; Yao, C.; Catani, F.; Zhou, C.; Huang, J. Uncertainties of landslide susceptibility prediction considering different landslide types. J. Rock Mech. Geotech. Eng. 2023, 15, 2954–2972. [Google Scholar] [CrossRef]
- Lei, T.; Zhang, Y.; Lv, Z.; Li, S.; Liu, S.; Nandi, A.K. Landslide Inventory Mapping from Bitemporal Images Using Deep Convolutional Neural Networks. IEEE Geosci. Remote Sens. Lett. 2019, 16, 982–986. [Google Scholar] [CrossRef]
- Althuwaynee, O.F.; Aydda, A.; Hwang, I.T.; Lee, Y.K.; Kim, S.W.; Park, H.J.; Lee, M.S.; Park, Y. Uncertainty reduction of unlabeled features in landslide inventory using machine learning t-SNE clustering and data mining apriori association rule algorithms. Appl. Sci. 2021, 11, 556. [Google Scholar] [CrossRef]
- Ramos-Bernal, R.N.; Vázquez-Jiménez, R.; Cantú-Ramírez, C.A.; Alarcón-Paredes, A.; Alonso-Silverio, G.A.; Bruzón, A.G.; Arrogante-Funes, F.; Martín-González, F.; Novillo, C.J.; Arrogante-Funes, P. Evaluation of conditioning factors of slope instability and continuous change maps in the generation of landslide inventory maps using machine learning (ML) algorithms. Remote Sens. 2021, 13, 4515. [Google Scholar] [CrossRef]
- Chen, X.; Zhao, C.; Lu, Z.; Xi, J. Landslide Inventory Mapping Based on Independent Component Analysis and UNet3+: A Case of Jiuzhaigou, China. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17, 2213–2223. [Google Scholar] [CrossRef]
- Senogles, A.; Olsen, M.J.; Leshchinsky, B. SlideSim: 3D Landslide Displacement Monitoring through a Physics-Based Simulation Approach to Self-Supervised Learning. Remote Sens. 2022, 14, 2644. [Google Scholar] [CrossRef]
- Zhou, H.; Dai, K.; Pirasteh, S.; Li, R.; Xiang, J.; Li, Z. InSAR Spatial-Heterogeneity Tropospheric Delay Correction in Steep Mountainous Areas Based on Deep Learning for Landslides Monitoring. IEEE Trans. Geosci. Remote Sens. 2023, 61, 5215014. [Google Scholar] [CrossRef]
- Liu, Y.; Xu, P.; Cao, C.; Zhang, W.; Zhao, M.; Zhu, K. Geomorphological transformations and future deformation estimations of a large potential landslide in the high-order position area of Diexi, China. Geocarto Int. 2023, 38, 2197514. [Google Scholar] [CrossRef]
- Ghaderpour, E.; Dadkhah, H.; Dabiri, H.; Bozzano, F.; Scarascia Mugnozza, G.; Mazzanti, P. Precipitation Time Series Analysis and Forecasting for Italian Regions. Eng. Proc. 2023, 39, 23. [Google Scholar] [CrossRef]
- Arrogante-Funes, P.; Bruzón, A.G.; Álvarez-Ripado, A.; Arrogante-Funes, F.; Martín-González, F.; Novillo, C.J. Assessment of the regeneration of landslides areas using unsupervised and supervised methods and explainable machine learning models. Landslides 2024, 21, 275–290. [Google Scholar] [CrossRef]
- Zheng, X.; He, G.; Wang, S.; Wang, Y.; Wang, G.; Yang, Z.; Yu, J.; Wang, N. Comparison of machine learning methods for potential active landslide hazards identification with multi-source data. ISPRS Int. J. Geoinf. 2021, 10, 253. [Google Scholar] [CrossRef]
- Gao, B.; He, Y.; Chen, X.; Zheng, X.; Zhang, L.; Zhang, Q.; Lu, J. Landslide Risk Evaluation in Shenzhen Based on Stacking Ensemble Learning and InSAR. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2023, 16, 1–18. [Google Scholar] [CrossRef]
- Maragaño-Carmona, G.; Fustos Toribio, I.J.; Descote, P.Y.; Robledo, L.F.; Villalobos, D.; Gatica, G. Rainfall-Induced Landslide Assessment under Different Precipitation Thresholds Using Remote Sensing Data: A Central Andes Case. Water. 2023, 15, 2514. [Google Scholar] [CrossRef]
- Chen, T.; Wang, Q.; Zhao, Z.; Liu, G.; Dou, J.; Plaza, A. LCFSTE: Landslide Conditioning Factors and Swin Transformer Ensemble for Landslide Susceptibility Assessment. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2024, 17, 6444–6454. [Google Scholar] [CrossRef]
- Shao, X.; Xu, C.; Li, L.; Yang, Z.; Yao, X.; Shao, B.; Liang, C.; Xue, Z.; Xu, X. Spatial analysis and hazard assessment of large-scale ancient landslides around the reservoir area of Wudongde hydropower station, China. Nat. Hazards 2024, 120, 87–105. [Google Scholar] [CrossRef]
- Qi, J.; Chen, H.; Chen, F. Extraction of landslide features in UAV remote sensing images based on machine vision and image enhancement technology. Neural. Comput. Appl. 2022, 34, 12283–12297. [Google Scholar] [CrossRef]
- Xia, W.; Chen, J.; Liu, J.; Ma, C.; Liu, W. Landslide extraction from high-resolution remote sensing imagery using fully convolutional spectral–topographic fusion network. Remote Sens. 2021, 13, 5116. [Google Scholar] [CrossRef]
- Yang, R.; Zhang, F.; Xia, J.; Wu, C. Landslide Extraction Using Mask R-CNN with Background-Enhancement Method. Remote Sens. 2022, 14, 2206. [Google Scholar] [CrossRef]
- Chen, H.; He, Y.; Zhang, L.; Yao, S.; Yang, W.; Fang, Y.; Liu, Y.; Gao, B. A landslide extraction method of channel attention mechanism U-Net network based on Sentinel-2A remote sensing images. Int. J. Digit. Earth 2023, 16, 552–577. [Google Scholar] [CrossRef]
Learning Algorithms | Prediction Accuracy | Data Sources | Reference |
---|---|---|---|
ANN | 0.97 | TERRASAR-X; ALOS–PALSAR | Masruroh et al. [51] |
CF–RF | 0.93 | ALOS–PALSAR; DEM; LISS-III; Radarsat | Peethambaran et al. [23] |
CF–RF, U-Net | 0.94; 0.96 | Landsat-8; DEM; ALOS–PALSAR | Sharma et al. [35] |
DCN | 0.97 | ALOS–PALSAR; Sentinel-2 | Aladhi et al. [36] |
ANN, RF | 0.97; 1.0 | ASTER-GDEM; Landsat-8 | Sun et al. [52] |
ANN, RF; SVM | 0.98; 0.95; 0.97 | PlanetScope; RapidEye; Sentinel-2 | Das et al. [53] |
MLP | 0.88 | DEM | Lee et al. [24] |
ANN, SVM, RF; ELM | 0.96; 0.96; 0.99; 0.99 | MODIS, DEM | Sharma et al. [41] |
RF, ANN, SVM | 0.94; 0.90; 0.90 | DEM; Landsat-8 | Hu et al. [54] |
FES, ELM | 0.84; 0.80 | ALOS–PALSAR; DEM; LISS-III; Radarsat | Peethambaran et al. [23] |
ANN, BN, LR, SVM | 0.92; 0.86; 0.93; 0.93 | ASTER-DEM | Xie et al. [55] |
DNN–HBA | 0.95 | DEM; Landsat-8 | Nguyen et al. [56] |
DADRCNN | 0.90 | DEM; OpenStreetMap | Ma et al. [57] |
CNN | 0.96 | SBAS–InSAR | Sheng et al. [58] |
FC–SAE | 0.85 | DEM; Landsat-8 | Huang et al. [33] |
ResNet-50 + SCAM | 0.99 | DEM | Ji et al. [59] |
ACO–DBN | 0.97 | DEM; GDCP | Xiong et al. [60] |
DNN; LSTM; CNN; RNN | 0.87; 0.87; 0.86; 0.83 | DEM; GEE | Habumugisha et al. [61] |
LD–BiLSTM | 0.90 | DEM | Wang et al. [62] |
GA–SVM | 0.96 | UAV; DEM | Niu et al. [63] |
RF, BRT, CART, GLM | 0.78; 0.96; 0.82; 0.82 | SRTM; DEM | Youssef et al. [29] |
LSSVM–BC | 0.90 | SPOT satellite images | Tien et al. [64] |
LR, LB, NB | 0.84; 0.71; 0.85 | DEM | Pourghasemi et al. [65] |
LNRF–BRT, LNRFLMR | 0.91; 0.90 | PALSAR; DEM | Arabameri et al. [66] |
RF | 0.97 | Sentinel-1 and -2 | Kocama et al. [67] |
ANN, RF, DT, SVM | 0.93; 0.92; 0.89; 0.91 | ALOS–PALSAR; DEM | AI-Najjar et al. [68] |
RF, XGBoost, NB, ANN, KNN, FR | 0.99; 0.99; 0.97; 0.92; 0.88; 0.67 | ALOSPALSAR; GDEM; Landsat-8 | Khalil et al. [69] |
Metric Measures | Parameters | Reference |
---|---|---|
True positive (TP), false negative (FN): (P) is the total number of landslides; and (N) is the total number of non-landslides. True negative (TN), false positive (FP), and Mathew coefficient correlation (MCC) | Hong et al. [15], Huang et al. [70], Nhu et al. [11], Hussain et al. [71], Sun et al. [52], Yuan et al. [39] | |
pobserved is identical to accuracy and pexpected | Lee et al. [24] | |
Precision (PRE) | Chowdhuri et al. [72], Zhang et al. [73] | |
Mean absolute error (MAE), p1 is the predicted value of landslide sensitivity, a1 is the actual value of landslide sensitivity, and n is the number of sample instances. | Deng et al. [74] | |
Mean Intersection over Union (mIoU), True positive (TP), false positive (TN), false negative (FN) | Liu et al. [34] |
No. of Landslide Factors | No. of Landslide Maps | Source | Algorithms | Results/Metric | Reference |
---|---|---|---|---|---|
9 | 226 | UAV, DEM | GA–SVM (best) and SVM | OA = 0.959, Kappa = 0.957 | Niu et al. [63] |
14 | 82 | Surface area ratio (SAR), DEM, Landsat/ETM+ satellite | SVM (best) | Polynomial degree-3 with AUC of 0.88 | Pourghasemi et al. [93] |
12 | 1 | Landsat, DEM | SVM15% (best), DT, LR | Kappa-index15% = 0.57, AUC15% = 0.79, FP15% = 0.4 | Marjanović et al. [13] |
11 | 125 | QuickBird, SRTM, DEM, Enhanced thematic mapper plus ETM+ | RF, BRT (best), CART, and GLM | AUC = 0.958, Precision rate = 0.86 | Youssef et al. [29] |
9 | 639 | DEM, Landsat-8 | SOM–ELM (best), SOM–SVM, and single ELM | Prediction rate 0.867, FPR = 10.2% | Huang et al. [2] |
8 | N/M | ASTER–GDEM | SVM, PSVM, and L2–SVM–MFN (best) | AUC = 0.83, AA = 0.72, TPR = 0.93, TNR = 0.72 | Kumar et al. [30] |
17 | N/M | Ultracam-X, DEM | RF and Boosted Tree (best) | Landslide susceptibility index rank-regression = 84.87 and classification = 85.98% | Kim et al. [14] |
16 | 249 | GEE, ALOS–PALSAR, DEM, Sentinal-2, Landsat-8 | ADTree, FLDA, and RF (best) | Susceptibility zone of = 11.35%, AUC = 0.97 | Arabameri et al. [49] |
20 | 151 | DEM | Multi-layer perceptron (MLP) | Sensitivity = 82.61%, accuracy = 80.43%, PPV = 79.17%, NPV = 81.82%, Kappa = 0.61 and AUC = 0.88 | Lee et al. [24] |
17 | 152 | InSAR, GEI, Sentinel-2, Sentinel-1, and Landsat-8 | AB (best), ADTree, and AB–ADTree | AUC = 0.96 | Nhu et al. [11] |
17 | 544 | DEM, GEE | Support Vector Classification (SVC) | Accuracy of 0.91 and AUC = 0.96 | Qing et al. [94] |
13 | 210 | SRTM, Landsat-7 and Landsat-8, GEE | LR (best), SE, WoE, and FR | AUC = 85.3%, precision = 83.1% | Ahmad et al. [95] |
11 | 35 | LiDAR, RIEGL, DEM | ANN with Ohe-X | AUC = 0.93 with an improvement rate of 37.15% | Al-Najjar et al. [96] |
19 | 616 | Resources satellite three (ZY-3), ASTER GDEM | ANN, LR, and SVM (best) | AUC = 0.93 with accuracy = 83.86% | Xie et al. [55] |
16 | 7184 | DEM, Geospatial Data Cloud Platform | ACO–DBN | Accuracy = 93%, precision = 95%, sensitivity = 91% and F1-score = 0.93 | Xiong et al. [60] |
12 | 7556 | Landsat-8, DEM, ALOS, GEE | SE–CapNet (best), CapNet, CNN, and RF | Accuracy = 96%, precision = 97%, sensitive = 95% and AUC = 0.97 | Chang et al. [97] |
10 | 13,886 | ALOS | BO_GBT, BO_RF, SVM, and CNN_BO_GBT (best) | Accuracy = 0.79, F1-score = 0.84, Kappa = 0.67 and precision = 0.82 | Gao et al. [26] |
13 | 1290 | GEE, DEM | CNN, DNN (best), LSTM, and RNN | F1-score = 0.82 and AUC = 0.873 | Habumugisha et al. [61] |
12 | 413 | DEM, Landsat-8 OLI | DRNN | AUC = 0.96 (train and tested), Kappa = 0.829 | Huang et al. [98] |
12 | 235 | ASTER GDEM, Landsat-8, DEM | SVM and RF (best) | AUC > 0.80 for landslide aggregate, neighborhood factor-based | Huang et al. [70] |
12 | 400 | SRTM, Landsat-8 OLI, DEM | RF, SVM, maxENT, GBM, and LR | SVM, RF, and GBM demonstrated the best results, but the AUC of SVM = 0.969 and performance overall (POA = 2669) | Shahzad et al. [99] |
9 | 213 | ALOS–PALSAR, DEM | FSP, FES, and ANN | AUC > 0.80 | Sweta et al. [89] |
19 | 59 | SBAS–InSAR, Setinel-2, Setinel-1, SRTM DEM, GEE | XGBoost | AUC = 0.996, accuracy = 97.98%, TPR = 98.77%, F1-score = 0.98, and Kappa = 0.96 | Wang et al. [100] |
13 | 1081 | ASTER GDEM, Landsat-8 OLI, DEM | CF–RF | POA = 257.046, AUC = 0.946 | Yuan et al. [39] |
12 | 233 | GEE, Landsat-8 satellite, ASTER GDEM | WLR, WLightGBM, and WRF (best) | Balance accuracy = 0.84, G-mean = 0.84, recall = 0.90, accuracy = 0.79 and AUC = 0.91 | Zhang et al. [101] |
11 | 405 | GDEMV2, Landsat-8 OLI, DEM | CNN | Precision = 0.99, recall = 0.91, F1-score = 0.94, MCC = 0.77, SRC-AUC = 93.14% and PRC-AUC = 91.81% | Zhang et al. [102] |
16 | 256 | DEM, Geographic Remote sensing Ecological network | SVM, LR, RF, XGBoost (best), and LDA | AUC = 0.876 | Cao et al. [103] |
15 | 235 | ASTER GDEM, Landsat-8 TM | Slope-RF and Slope-MLP | Both models performed equally with AUC = 0.895 | Chang et al. [104] |
14 | 293 | Landsat-8, SRTM DEM, GEE | SVM, RF, and GAMI-net (best) | AUC = 0.94, accuracy = 0.87, precision = 0.86, F1-score = 0.87, and recall = 0.87 | Fang et al. [76] |
17 | 85 | UAV, GEI, ALOS–PALSAR, TERRASAR-X | ANN | AUC = 0.965 and precision-recall = 0.976 | Masruroh et al. [51] |
12 | 30 | Landsat-8, DEM | ANN, boosting-C5.0 DT, and SVM | AUC = 0.99 | Miao et al. [105] |
22 | 86 | GEE, Landsat-9, DEM, JAXA ALOSWorld 3D | RF and XGBoost | Both models achieved AUC = 0.96 | Parra et al. [106] |
13 | 2308 | DEM, Landsat-8, ALOS | LR, RF (best), NB, and MLP | AUC = 0.92, susceptibility = 12.24% | Wang et al. [107] |
13 | 302 | Sentinel-2, Landsat-8, SRTM, DEM, GEI | RF, EXT–XGBoost (best), and KNN | AUC = 0.97, and all models have an accuracy rate > 90% | Abbas et al. [108] |
12 | 1215 | DEM, Maps | SVM, LR, LDA, RF, and XGBoost | AUC > 90% for all the models | Agboola et al. [83] |
15 | 811 | DEM | ABC–ANFIS | Precision = 0.89, AUC = 0.96, RMSE = 0.21, MAE = 0.21, and Kappa = 0.70 | Chen et al. [109] |
16 | 337 | ASTER GDEM, Landsat-8 | MLP and RF | AUC > 0.9 for both models | Huang et al. [110] |
16 | 370 | DEM, Landsat-8 TM | SVM, MLP, and RF (best) | AUC > 0.90, but the accuracy decreases with increasing random error | Huang et al. [111] |
11 | 407 | ASTER V2 and V3, COP-DEM, ALOS, FABDEM, SRTM | KNN and RF (best) | AUC > 0.80 and 0.95 COP | Lu et al. [112] |
14 | 945 | DEM, Landsat-8 | DNN and DNN-hybrid models (DNN–MPA) (best) | AUC = 0.96, with accuracy > 90% | Nguyen et al. [56] |
14 | 170 | GEI, DEM, Sentinel-2 | SVM | AUC = 0.88, recall = 0.73, precision = 0.86, accuracy = 0.81, and F1-score = 0.79 | Patil et al. [84] |
16 | 489 | MERIT DEM, SRTM DEM, GEE, MODIS | SVMSMOTE | Accuracy = 96%, sensitivity = 97% and MCC = 0.92 | Sharma et al. [41] |
12 | 12 | Landsat-8 OLI, DEM, GEE, ASTER GDEM | LGBM (best), GRU, LSTM, RNN, RF, and ET | F1-score = 0.56, recall = 0.52, precision = 0.60, AUC = 0.84, and accuracy = 0.82 | Song et al. [42] |
15 | 1718 | Landsat-8 OLI, DEM | RF (best), and XGBoost | AUC > 0.80 | Sun et al. [113] |
13 | 78,529 | MT–InSAR, DEM | Stacking RF-XGBoost | AUC = 0.96 | Zeng et al. [114] |
9 | 202 | GEE, Landsat-5 TM, Landsat-8 OLI | CTLGNet | AUC = 0.97 and recall = 0.98 | Zhao et al. [115] |
Data Used | Algorithms Evaluated | Some Key Results | Reference |
---|---|---|---|
DEM | Deep convolution Neural Network (DCNN) | The DCNN model demonstrates a high capacity for detecting landslides in mountainous terrain. | Chen et al. [22] |
Unmanned aerial vehicles (UAVs), DEM | CNN models | The CNN model shows a precision equivalent to 0.90, an F-score of 0.85, and mIoU of 0.74; the effectiveness of CNN for slope or landslide detection depends strongly on its design. CNN proves to be a powerful tool, but it depends on trial and error. | Ghorbanzadeh et al. [78] |
DEM, satellite images | ResNet-50 | The ResNet-50 achieved a precision as high as 0.987, a recall of 0.9455, an accuracy of 0.9816, and an F1-score of 0.9662, indicating a successful prospect of landslide detection with satellite images. | Ji et al. [59] |
GEE | You only look once-small attention (YOLO-SA) | Comparing the YOLO-SA model to the other 11 models, the YOLO-SA has an accuracy of 0.94 and an F1-score of 0.91, showing a potential landslide detection. | Cheng et al. [142] |
Sentinel-2A, DEM, ALOS–PALSAR, GEE | convolutional auto-encoder (CAE) | The results show that CAE with a mini-batch K-means clustering algorithm can be applied to primary landslide mapping. | Shahabi et al. [87] |
Zeiss RMK TOP 15 aerial survey, DEM | Object-oriented change detection CNN (CDCNN) | The CDCNN shows a better and more robust detection of landslides in vast areas, even with a considerable volume of datasets and complex land cover, requiring less human interaction. | Shi et al. [143] |
Landsat-8 and JAXA ALOS DSM and (GEE) | SVM, CART, minimum distance, RF and NB | The SVM and RF models achieved the best detection results with a true positive ratio of 87.5% compared to others. | Singh et al. [144] |
UAV images | ResNet-50 and ResNet-101 by exploring Mask R-CNN | With the incorporation of Mask R-CNN, the ResNet-101 performed better than ResNet-50 with a precision of 1.0, recall of 0.93, and F1-score of 0.97, indicating good landslide detection. | Ullo et al. [145] |
PlanetScope, RapidEye, Sentinel-2, ASTER, GDEM and DEM | ANN, RF, and SVM | From the studies, the results from all three models using remote sensing datasets provided high accuracies of 0.94, sensitivity of 0.96, and specificity of 0.92, demonstrating the models’ robustness in accurately detecting landslides. | Das et al. [53] |
Sentinel-2, ALOS DEM | ResU-Net, OBIA, and ResU-Net-OBIA | According to the proposed model ResU-Net-OBIA, the model detects landslide with a precision of 0.73, recall of 0.80, and F1-score of 0.77 | Ghorbanzdeh et al. [18] |
GEE, Sentinel-2, Landsat-8 | Transfer learning Mask R-CNN (TL-Mask R-CNN) | The result shows that the TL-Mask R-CNN model can detect landslides with a recall of 0.78 and an F1-score of 0.80, demonstrating a solid detection and segmentation along Sichuan–Tibet transportation and landslide recognition along the Sichuan–Tibet transportation corridors. | Jiang et al. [146] |
Sentinel-2, GEE, DEM and ALOS PALSAR | DeepLab-v2, DeepLab-v3+, FCN-8s, LinkNet, FRRN-A, FRRN-B, SQNet, U-Net, and ResU-Net. | The ResU-Net performed better for detecting landslides than all the other models after all the models were trained from scratch on patches. | Ghorbanzdeh et al. [147] |
RapidEye, DEM and ALOS–PALSAR | U-Net, SVM, KNN and RF | The U-Net model achieved slightly better results than the other learning algorithms; however, it is still at the preliminary stage of detecting landslides. | Meena et al. [148] |
Airborne remote sensing image | HRNet, DeepLabV3, Attention-UNet, U2Net, FastSCNN and SegFormer | Comparing the experimental results, the SegFormer model enhances landslide detection accuracy of the mIoU by 2.2%, IoU by 5%, and F1-score by 3% and reduces the pixel-wise classification error ratio by 14% | Tang et al. [149] |
UAV | Transfer learning (TL) model | The result from UAV remote sensing images with a proposed TL model demonstrated a better detection of landslides when accurate landslide data are used. | Yang et al. [50] |
TripleSat satellite and the RGB images | U-Net, DeepLab v3+ and PSPNet | After experimenting with the three models with different backbone networks, PSNet with ResNet-50 demonstrated landslide detection with an mIoU of 91%, a recall of 97%, and a precision of 94% | Yang et al. [150] |
RapidEye, Geoeye-1, GEE, TripleSat, Sentinel-2 and GaoFen-2 | HADeenNet | With ResNet50 as the backbone, HADeenNet modes built with the DL framework increase the F1-score by about 21% and 10% higher than six traditional deep learning models. | Yu et al. [151] |
TripleSat | U-Net with ResNet-50, ResNet-101, VGG-19, and DenseNet-121 as backbone | Comparing the U-Net model with the four backbones, the U-Net + ResNet50 demonstrates better landslide detection by recording the highest precision value of 0.98, a recall of 0.98, an F1-score of 0.98, an overall accuracy (OA) of 1.0, and an MCC of 0.88. Although, the other three models show measured accuracy values > 0.90. | Chandra et al. [152] |
Google Earth Pro | SG-FCNN + MCMO | The results from the study show that the SG-FCNN+MCMO can detect landslides and reduce the false and miss detection of landslides compared to the traditional FCNN model. | Chen et al. [137] |
Sentinel-2, and ALOS PALSAR DEMs | U-Net | The U-Net model demonstrates a promise for case event inventory but shows lower detection accuracy for geomorphology inventory. | Das et al. [80] |
CBERS-04A and remote sensing | CNN, Unet, FPN and Linknet | The models show recall values similar for all (>75%) with precision values < 20%, while fewer precision values were attributed to the false positive sample. Additionally, the models cannot detect relict landslides in terrains with rainforests. | Garcia et al. [92] |
Optically sensing | Faster R-CNN with VGG16 and ResNet50 as backbone | The Faster R-CNN+ResNet50 performed better than the Faster R-CNN+VGG16 with an AP of about 91%, an F1-score of 94%, a recall of 91%, and a precision of 98% and can detect slope failure and landslides, while R-CNN+VGG16 can detect small-slope failure and landslides. | Guan et al. [153] |
LiDAR and optical images, DEM | ResUNet, LandsNet, HRNet, MLP, SegFormer, and proposed (DemDet) | The DemDet model has a mean accuracy of 0.95, mIoU of 0.61, and F1-score of 0.78, demonstrating the capability of the model to detect landslide in a forest cover compared to the ResUNet, LandsNet, HRNet, MLP, and SegFormer models | Li et al. [154] |
Shipborne images | Transfer learning models and VGG19, DenseNet121, EfficientNetB0, SEResNet50, and ViT | The result shows that the decision-level fusion and transfer learning integration can classify landslide-based shipborne images. | Li et al. [19] |
TripleSat satellite | Vision transformer (ViT) models | Comparing the ViT model with the traditional deep learning models, the ViT model has the benefit of detecting landslides with remote-sensing images. | Lv et al. [155] |
Landsat-8, GEE | Mask RCNN + weight transfer function + Mask IoU model | The proposed model precision is enhanced by 20% compared to Mask R-CNN, and the average accuracy higher than 0.75 achieved for the proposed mask IoU threshold is 0.5. | Wang et al. [156] |
Sentinel-2 Level 1C, DEM and ALOS DEM | Semantic segmentation network (EGCN) | The EGCN model outperformed traditional major deep learning models by obtaining an OA of 0.998, an mIoU of 0.997, a Kappa of 0.973, an F1-score of 0.974, a precision of 0.973, and a recall of 0.974, indicating the model’s capability of recognized landslides. | Yang et al. [157] |
Synthetic aperture radar (SAR), Sentinel-1A, Sentinel-2A and DEM | Multi-input channel U-Net | The U-Net with multi-input channel (16) demonstrates a good prospect for detecting landslide with a recall value of 69.74%, precision of 62.96%, F1-score of 66.18, and an mIoU of 74.67% | Chen et al. [158] |
Sentinel-2 | U-Net, U2-Net, and U-Net3+ | The U-Net models applied provided the best results for detecting traces of landslides with remote sensing data. | Dang et al. [159] |
Landslide Factors | Data Used | Algorithms Evaluated | Results/Metric | Reference |
---|---|---|---|---|
Topography, hydrological, geological, land cover | Landsat TM8, DEM | FC–SAE, SVM and BPNN | Prediction rate and total accuracy of FC–SAE (0.85), SVM (0.82 and 0.81), and BPNN (0.82 and 0.81); the FC–SAE algorithms proposed predict the landslide susceptibility better compared to the other models. | Huang et al. [33] |
Slope, aspect, elevation, lithology, rainfall, curvature | GEI, Landsat-8, DEM | MBNBT, MLPN, SVM, and NBT | AUC of MBNBT (0.82), MLPN (0.80), SVM (0.80), and NBT (0.80); MBNBT offers better prediction than the other learning algorithms | Nguyen et al. [27] |
TWI, NDVI, elevation, slope, aspect, curvature | GEE, Landsat-8, ASTER-GDEM | LR–GBDT–VFI | The proposed LR–GBDT–VFI model (AUC: 0.98) best predicts landslide potential after comparing the model with nine traditional models. | Sachdeva et al. [165] |
Slope, lithology, aspect, total curvature, STI | ALOS PALSAR, DEM | GAN and SMOTE for correcting data imbalanced with ANN, RF, DT, KNN and SVM | The results from the study show that landslide data balancing could affect the prediction capacity of learning algorithms | AI-Najjar et al. [68] |
Landslide, terrain, lithology, land cover, NDVI, landform and precipitation | DEM, Landsat, STS Endeavour OV-105, Crowdsourcing | Multi-graded Cascade Forest, thus, (GCF) algorithm | The GCF model shows a precision efficiency of 0.93 for landslide prediction, and the workflow alleviated poor prediction problems from limited landslide data. | Chen et al. [166] |
Rainfall, slope, curvature, TWI, soil texture, lineament density, NDVI | ALOS PALSAR, DEM, Sentinel-2 MSI, | RF and DNN | The RF and DNN models achieved higher accuracy of AUC greater than 0.90, which indicates a high prediction of landslide susceptibility, but DNN outperformed the RF with an AUC of 0.96. Rainfall was noticed to be a significant contributing factor to landslides, according to the results | Dahim et al. [40] |
Aspect, elevation, lithology, distance to a river | ASTER, DEM, Landsat-8 | MLP, RF, and proposed semi-supervised machine learning (SSML); MLP, RF, SVM, and low-pass filter method | The proposed SSML model reduced landslide position errors and significantly improved landslide accuracy compared to MLP and RF. The random errors in landslide condition factors result in higher uncertainty, and the low-pass filter method reduced random error significantly. | Huang et al. [110], Huang et al. [111] |
Slope, aspect, elevation, rainfall, TWI, NDVI, land cover | Global Digital Elevation Model (GDEM), ALOSPALSAR and Landsat 8 | RF, XGBoost, NB, ANN, KNN and FR | The AUC of RF (0.992), XGBoost (0.991), NB (0.970), ANN (0.922), KNN (0.877), and FR (0.674), indicating that the RF model provided effective means of predicting landslide on a global scale, followed by XGBoost and ANN comparing the models. | Khalil et al. [69] |
Slope, aspect, elevation, sediment transport index, lithology, fault proximity, TWI, NDVI | Resources-1 multispectral LISS-IV, IRS-P5 Cartosat-1, Landsat-7, ALOS–PALSAR, DEM, TRMM, and Google Earth Engine | Multilayer Perceptron (MLP) | The success and precision rate curves under AUC are 0.94 and 0.92, respectively, indicating high prediction accuracy of MLP for forecasting landslide potentials. | Sundriya et al. [167] |
Length-slope, topography, vegetation cover, lithology, rainfall, TWI | DEM | Landslide density-based bidirectional long short-term memory (LD–BiLSTM) model compared with Landslide object model (LO–BiLSTM model) | The precision, recall F1-score, and AUC are 0.90, 0.89, 0.90, and 0.94 for LD–BiLSTM, respectively, and 0.81, 0.82, 0.81, and 0.91 for LO–BiLSTM, respectively, making LD–BiLSTM model superior to predict landslide compared to LO–BiLSTM. | Wang et al. [62] |
Algorithms | Source | Objective | Landslide Factors | Accuracy Evaluation Methods | Some Key Result | Reference |
---|---|---|---|---|---|---|
LR, NB, and LB | DEM and other multiple-source | Landslide susceptibility assessment | Seventeen factors were used, including aspect, slope, topographic wetness index (TWI), lithology, land used/land cover | Area under curve (AUC)-LR = 84.2%, NB = 70.7% and LB = 85% | LR and LB reveal a reasonable accuracy for landslide susceptibility assessment than NB | Pourghasemi et al. [65] |
ANN, GBM, and MaxEnt | InSAR, LiDAR | Slow movement landslide risk assessment | Slope angle, aspect, road density, TWI, TPI and stream density | Area under receiver operating characteristics (AUROC) = 0.96 and true skill statistics (TSS) = 0.82 | The models present reasonable results to assess landslide risk with remote-sensing data | Novellino et al. [28] |
RF and GradientBoost (GB) | DEM, Gaofen-1 satellite, GEI, etc. | Factor correlation analysis of landslide | Six factors from human activities, geological structures, geomorphology, and mineral composition | Receiver operating characteristics (ROC) for RF and GB = 0.97, and accuracy for RF and GB are 0.92 and 0.96, respectively. | RF and GB offer a better assessment of landslides in a complex geomorphic and tectonic terrain. NDVI was revealed to be closely associated with landslides compared to the other factors. | Qi et al. [17] |
NB, DT, SVM, and RF | InASR and another multi-data source | Potential landslide hazards assessment | Seventeen factors, including NDVI, slope, degree of weathering | Precision for NB = 0.66, DT = 0.76, RF = 0.77 and SVM = 0.87, Recall for NB = 0.66, DT = 0.77, RF = 0.77 and SVM = 0.83, F1 for NB = 0.66, DT = 0.75, RF = 0.74 and SVM = 0.82 | Among the four different models, the SVM provides the best prediction result | Zheng et al. [180] |
Fuzzy set procedure (FSP), fuzzy expert system (FES), and ANN | ALOS–PALSAR, DEM | Landslide susceptibility zoning assessment | Nine factors like slope, aspect, elevation, TWI, land use/landcover (LULC), curvature, lithology | AUC for FSP = 0.78, FES = 0.83 and ANN = 0.90 | The ANN was found to provide superior accuracy in landslide mapping | Sweta et al. [89] |
Stacking ensemble learning (SEL) | InSAR | Landslide risk evaluation | Topography, human activities, normalized difference vegetation index (NDVI) | ROC curve shows 8% accuracy | The SEL model provides the best evaluation accuracy for landslide mapping | Gao et al. [181] |
Logic model (LM) | Shuttle Radar Topography Mission (SRTM) | Rainfall-induced landslide assessment | Hydrology and rainfall, topography | From the ROC curve, the AUC of LM = 0.79 | The result indicated that the LM could map landslide-prone terrain based on precipitation and geomorphology. | Maragano et al. [182] |
LR, RF, NB, and multi-layer perception (MLP) | DEM, ALOS, and landsat-8 | Susceptibility assessment of landslide | Surface cover, human activities, hydrology, lithology, topography | AUC of LR = 0.79, RF = 0.92, NB = 0.79 and MLP = 0.84 | All the models show a good result, but the RF has the highest accuracy for landslide susceptibility assessment | Wang et al. [107] |
XGBoost | MODIS, DEM, ASTGTM3 | Landslide hazard assessment | Slope, climate, landcover, and geological | The AUC > 0.9 | The XGBoost model provided better performance and avoided overestimation of landslide susceptibility. | Zhang et al. [90] |
Landslide conditioning factor and swin transformer ensemble (LCFSTE) | GEE, Landsat-8 OLI | Landslide susceptibility assessment | Eleven factors, including slope, elevation, lithology, NDVI | AUC of LCFSTE = 0.94 compared to seven evaluation metrics | The LCFSTE model indicates a promising capability for assessing landslide susceptibility | Chen et al. [183] |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Akosah, S.; Gratchev, I.; Kim, D.-H.; Ohn, S.-Y. Application of Artificial Intelligence and Remote Sensing for Landslide Detection and Prediction: Systematic Review. Remote Sens. 2024, 16, 2947. https://doi.org/10.3390/rs16162947
Akosah S, Gratchev I, Kim D-H, Ohn S-Y. Application of Artificial Intelligence and Remote Sensing for Landslide Detection and Prediction: Systematic Review. Remote Sensing. 2024; 16(16):2947. https://doi.org/10.3390/rs16162947
Chicago/Turabian StyleAkosah, Stephen, Ivan Gratchev, Dong-Hyun Kim, and Syng-Yup Ohn. 2024. "Application of Artificial Intelligence and Remote Sensing for Landslide Detection and Prediction: Systematic Review" Remote Sensing 16, no. 16: 2947. https://doi.org/10.3390/rs16162947
APA StyleAkosah, S., Gratchev, I., Kim, D.-H., & Ohn, S.-Y. (2024). Application of Artificial Intelligence and Remote Sensing for Landslide Detection and Prediction: Systematic Review. Remote Sensing, 16(16), 2947. https://doi.org/10.3390/rs16162947