Improving Spatial Agreement in Machine Learning-Based Landslide Susceptibility Mapping

A review on landslide susceptibility mapping research in Bangladesh

Landslide susceptibility mapping is a common practice for landslide susceptibility assessment across the world. Like many other mountainous areas of the world, Bangladesh is facing frequent catastrophic landslides causing severe damage to the economy and society. As a result, several types of research have been conducted on landslides in Bangladesh. In the current research, a systematic review is conducted on the existing literature related to landslide susceptibility mapping to assess its contemporary trend with global research. The publications analyzed in this research were extracted from a website comprising landslide research of Bangladesh and by manual search. The aspects of the literature considered are year of publication, the journal where published, location/size of the study area, landslide inventory data type, susceptibility assessment/mapping method, thematic variables used, DEM characteristics, accuracy assessment methods and acquired accuracy of the models. The Chi-square test was conducted and correlation was measured to assess relation between selected features and map accuracy but no significant relationship was found. The studies are concentrated into three administrative districts of Chattogram, Rangamati and Cox's Bazar mainly covering the city centre. The publication rate is increasing but not following the global trend. Though various types of models are used and compared, the application of machine and deep learning algorithms are very limited and no evidence of Physically-based methods is found. Most of the cases, landslide inventory is prepared by conducting field survey, but the size is small. The research will help future practitioner in landslide susceptibility mapping research in the area.

Landslide Monitoring along the Dadu River in Sichuan Based on Sentinel-1 Multi-Temporal InSAR

The Dadu River travels in the mountainous areas of southwestern China, one of regions with the most hazards that has long suffered from frequent geohazards. The early identification of landslides in this region is urgently needed, especially after the recent Luding earthquake (MS 6.8). While conventional ground-based monitoring techniques are limited by the complex terrain conditions in these alpine valley regions, space interferometric synthetic aperture radar (InSAR) provides an incomparable advantage in obtaining surface deformation with high precision and over a wide area, which is very useful for long-term and slow geohazard monitoring. In this study, more than 500 Sentinel-1 SAR images with four frames acquired during 2017~2022 were collected to detect the hidden landslide regions from the Jinchuan to Ebian Section along the Dadu River, based on joint-scatterer InSAR (JS-InSAR) and small baseline subset (SBAS) techniques. The results showed that our method could be successfully applied for landslide monitoring in complex mountainous regions. Furthermore, 143 potential landslide regions spreading over an 800 km area along the Dadu River were extracted by integrating the deformation measurements and optical images. Our study can provide a reference for large-scale geological hazard surveys in mountainous areas, and the InSAR technique will be encouraged for the local government in future long-term monitoring applications in the Dadu River Basin.

An objective absence data sampling method for landslide susceptibility mapping

The accuracy and quality of the landslide susceptibility map depend on the available landslide locations and the sampling strategy for absence data (non-landslide locations). In this study, we propose an objective method to determine the critical value for sampling absence data based on Mahalanobis distances (MD). We demonstrate this method on landslide susceptibility mapping of three subdistricts (Upazilas) of the Rangamati district, Bangladesh, and compare the results with the landslide susceptibility map produced based on the slope-based absence data sampling method. Using the 15 landslide causal factors, including slope, aspect, and plan curvature, we first determine the critical value of 23.69 based on the Chi-square distribution with 14 degrees of freedom. This critical value was then used to determine the sampling space for 261 random absence data. In comparison, we chose another set of the absence data based on a slope threshold of < 3°. The landslide susceptibility maps were then generated using the random forest model. The Receiver Operating Characteristic (ROC) curves and the Kappa index were used for accuracy assessment, while the Seed Cell Area Index (SCAI) was used for consistency assessment. The landslide susceptibility map produced using our proposed method has relatively high model fitting (0.87), prediction (0.85), and Kappa values (0.77). Even though the landslide susceptibility map produced by the slope-based sampling also has relatively high accuracy, the SCAI values suggest lower consistency. Furthermore, slope-based sampling is highly subjective; therefore, we recommend using MD -based absence data sampling for landslide susceptibility mapping.

A novel framework for addressing uncertainties in machine learning-based geospatial approaches for flood prediction

Globally, many studies on machine learning (ML)-based flood susceptibility modeling have been carried out in recent years. While majority of those models produce reasonably accurate flood predictions, the outcomes are subject to uncertainty since flood susceptibility models (FSMs) may produce varying spatial predictions. However, there have not been many attempts to address these uncertainties because identifying spatial agreement in flood projections is a complex process. This study presents a framework for reducing spatial disagreement among four standalone and hybridized ML-based FSMs: random forest (RF), k-nearest neighbor (KNN), multilayer perceptron (MLP), and hybridized genetic algorithm-gaussian radial basis function-support vector regression (GA-RBF-SVR). Besides, an optimized model was developed combining the outcomes of those four models. The southwest coastal region of Bangladesh was selected as the case area. A comparable percentage of flood potential area (approximately 60% of the total land areas) was produced by all ML-based models. Despite achieving high prediction accuracy, spatial discrepancy in the model outcomes was observed, with pixel-wise correlation coefficients across different models ranging from 0.62 to 0.91. The optimized model exhibited high prediction accuracy and improved spatial agreement by reducing the number of classification errors. The framework presented in this study might aid in the formulation of risk-based development plans and enhancement of current early warning systems.

Application of Transformer Models to Landslide Susceptibility Mapping

Landslide susceptibility mapping (LSM) is of great significance for the identification and prevention of geological hazards. LSM is based on convolutional neural networks (CNNs); CNNs use fixed convolutional kernels, focus more on local information and do not retain spatial information. This is a property of the CNN itself, resulting in low accuracy of LSM. Based on the above problems, we use Vision Transformer (ViT) and its derivative model Swin Transformer (Swin) to conduct LSM for the selected study area. Machine learning and a CNN model are used for comparison. Fourier transform amplitude, feature similarity and other indicators were used to compare and analyze the difference in the results. The results show that the Swin model has the best accuracy, F1-score and AUC. The results of LSM are combined with landslide points, faults and other data analysis; the ViT model results are the most consistent with the actual situation, showing the strongest generalization ability. In this paper, we believe that the advantages of ViT and its derived models in global feature extraction ensure that ViT is more accurate than CNN and machine learning in predicting landslide probability in the study area.

Dynamic landslide susceptibility analysis that combines rainfall period, accumulated rainfall, and geospatial information

Worldwide, catastrophic landslides are occurring as a result of abnormal climatic conditions. Since a landslide is caused by a combination of the triggers of rainfall and the vulnerability of spatial information, a study that can suggest a method to analyze the complex relationship between the two factors is required. In this study, the relationship between complex factors (rainfall period, accumulated rainfall, and spatial information characteristics) was designed as a system dynamics model as variables to check the possibility of occurrence of vulnerable areas according to the rainfall characteristics that change in real-time. In contrast to the current way of predicting the collapse time by analysing rainfall data, the developed model can set the precipitation period during rainfall. By setting the induced rainfall period, the researcher can then assess the susceptibility of the landslide-vulnerable area. Further, because the geospatial information features and rainfall data for the 672 h before the landslide's occurrence were combined, the results of the susceptibility analysis could be determined for each topographical characteristic according to the rainfall period and cumulative rainfall change. Third, by adjusting the General cumulative rainfall period (D_G) and Inter-event time definition (IETD), the preceding rainfall period can be adjusted, and desired results can be obtained. An analysis method that can solve complex relationships can contribute to the prediction of landslide warning times and expected occurrence locations.

Spatiotemporal distribution characteristics, causes, and prevention advice of fatal geohazards in Jiangxi Province, China

Fatal geohazards result in severe losses of life and property worldwide, thus urging many large-scale studies of such geohazards. Further research on hotspots prone to fatal geohazard identified in national-scale studies is critical for government geohazard prevention. It has been pointed out that more detailed small-scale (sub-national) studies are essential for the hotspots (e.g., Jiangxi Province) identified in national-scale studies. However, there are only a few small-scale studies of hotspots and earlier studies have rarely delved into a thorough and detailed analysis of hotspots. In addition, previous studies of fatal geohazards have failed to offer specific geohazard prevention advice, significant for geohazard control policies. To bridge these gaps, this study took advantage of the Jiangxi Inventory of Fatal Geohazards (JIFGH) and employed spatial analysis and the geographical detector to analyze the spatiotemporal characteristics and causes and present prevention advice on fatal geohazards in Jiangxi Province. The study also analyzes the importance of provincial-scale (first-level administrative scale) studies for hotspots identified in national-scale studies. JIFGH includes 386 non-seismically triggered fatal geohazards that caused a total of 979 fatalities in the 1960-2020 period. The temporal trend of fatal geohazards in Jiangxi Province is mainly affected by rainfall and the government geohazard prevention measures. The causes of most fatal geohazards in Jiangxi Province include (i) slope-cutting activities in house construction projects that create steep slopes prone to failure, which threaten the vulnerable residents and buildings nearby and (ii) rainfall that triggers failures of cut slopes. This study not only proposes geohazard prevention advice for Jiangxi Province and tectonically stable areas but also analyzes the significance of provincial-scale studies of hotspots identified in national-scale studies. Therefore, this study contributes to the prevention of fatal geohazards in Jiangxi Province and tectonically stable areas, while also providing an essential reference for other studies of fatal geohazards.

Landslide Susceptibility Mapping Using Machine Learning: A Literature Survey

Landslide is a devastating natural disaster, causing loss of life and property. It is likely to occur more frequently due to increasing urbanization, deforestation, and climate change. Landslide susceptibility mapping is vital to safeguard life and property. This article surveys machine learning (ML) models used for landslide susceptibility mapping to understand the current trend by analyzing published articles based on the ML models, landslide causative factors (LCFs), study location, datasets, evaluation methods, and model performance. Existing literature considered in this comprehensive survey is systematically selected using the ROSES protocol. The trend indicates a growing interest in the field. The choice of LCFs depends on data availability and case study location; China is the most studied location, and area under the receiver operating characteristic curve (AUC) is considered the best evaluation metric. Many ML models have achieved an AUC value > 0.90, indicating high reliability of the susceptibility map generated. This paper also discusses the recently developed hybrid, ensemble, and deep learning (DL) models in landslide susceptibility mapping. Generally, hybrid, ensemble, and DL models outperform conventional ML models. Based on the survey, a few recommendations and future works which may help the new researchers in the field are also presented.

Landslide Detection Based on ResU-Net with Transformer and CBAM Embedded: Two Examples with Geologically Different Environments

An efficient method of landslide detection can provide basic scientific data for emergency command and landslide susceptibility mapping. Compared to a traditional landslide detection approach, convolutional neural networks (CNN) have been proven to have powerful capabilities in reducing the time consumed for selecting the appropriate features for landslides. Currently, the success of transformers in natural language processing (NLP) demonstrates the strength of self-attention in global semantic information acquisition. How to effectively integrate transformers into CNN, alleviate the limitation of the receptive field, and improve the model generation are hot topics in remote sensing image processing based on deep learning (DL). Inspired by the vision transformer (ViT), this paper first attempts to integrate a transformer into ResU-Net for landslide detection tasks with small datasets, aiming to enhance the network ability in modelling the global context of feature maps and drive the model to recognize landslides with a small dataset. Besides, a spatial and channel attention module was introduced into the decoder to effectually suppress the noise in the feature maps from the convolution and transformer. By selecting two landslide datasets with different geological characteristics, the feasibility of the proposed model was validated. Finally, the standard ResU-Net was chosen as the benchmark to evaluate the proposed model rationality. The results indicated that the proposed model obtained the highest mIoU and F1-score in both datasets, demonstrating that the ResU-Net with a transformer embedded can be used as a robust landslide detection method and thus realize the generation of accurate regional landslide inventory and emergency rescue.

Landslide Susceptibility Assessment by Using Convolutional Neural Network

This study performs a GIS-based landslide susceptibility assessment using a convolutional neural network, CNN, in a study area of the Gorzineh-khil region, northeastern Iran. For this assessment, a 15-layered CNN was programmed in the Python high-level language for susceptibility mapping. In this regard, as far as the landside triggering factors are concerned, it was concluded that the geomorphologic/topographic parameters (i.e., slope curvature, topographical elevation, slope aspect, and weathering) and water condition parameters (hydrological gradient, drainage pattern, and flow gradient) are the main triggering factors. These factors provided the landside dataset, which was input to the CNN. We used 80% of the dataset for training and the remaining 20% for testing to prepare the landslide susceptibility map of the study area. In order to cross-validate the resulting map, a loss function, and common classifiers were considered: support vector machines, SVM, k-nearest neighbor, k-NN, and decision tree, DT. An evaluation of the results of the susceptibility assessment revealed that the CNN led the other classes in terms of 79.0% accuracy, 73.0% precision, 75.0% recall, and 77.0% f1-score, and, hence, provided better accuracy and the least computational error when compared to the other models.

Landslide Susceptibility Mapping Based on the Germinal Center Optimization Algorithm and Support Vector Classification

A landslide susceptibility model based on a metaheuristic optimization algorithm (germinal center optimization (GCO)) and support vector classification (SVC) is proposed and applied to landslide susceptibility mapping in the Three Gorges Reservoir area in this paper. The proposed GCO-SVC model was constructed via the following steps: First, data on 11 influencing factors and 292 landslide polygons were collected to establish the spatial database. Then, after the influencing factors were subjected to multicollinearity analysis, the data were randomly divided into training and testing sets at a ratio of 7:3. Next, the SVC model with 5-fold cross-validation was optimized by hyperparameter space search using GCO to obtain the optimal hyperparameters, and then the best model was constructed based on the optimal hyperparameters and training set. Finally, the best model acquired by GCO-SVC was applied for landslide susceptibility mapping (LSM), and its performance was compared with that of 6 popular models. The proposed GCO-SVC model achieved better performance (0.9425) than the genetic algorithm support vector classification (GA-SVC; 0.9371), grid search optimized support vector classification (GRID-SVC; 0.9198), random forest (RF; 0.9085), artificial neural network (ANN; 0.9075), K-nearest neighbor (KNN; 0.8976), and decision tree (DT; 0.8914) models in terms of the area under the receiver operating characteristic curve (AUC), and the trends of the other metrics were consistent with that of the AUC. Therefore, the proposed GCO-SVC model has some advantages in LSM and may be worth promoting for wide use.

Automatic Extraction of Potential Landslides by Integrating an Optical Remote Sensing Image with an InSAR-Derived Deformation Map

Landslide extraction is one of the most popular topics in remote sensing. Numerous techniques have been proposed to manage the landslide identification problem. However, most aim to extract landslides that have already occurred or delineate the potential landslide manually. It is greatly important to identify and delineate potential landslides automatically, which has not been investigated. In this paper, we propose an automatic identification and delineation method, i.e., object-based image analysis (OBIA) of potential landslides by integrating optical imagery with a deformation map. We applied a deformation map generated by the interferometric synthetic aperture radar (InSAR) technique, rather than the digital elevation model (DEM) for landslide segmentation. Then, we used a classification and regression tree (CART) model with the spectral, spatial, contextual and deformation characteristics for landslide classification. For accuracy assessment, we implemented the evaluation indicators of recall and precision. The proposed method is verified in both specific landslide-prone regions (Jinpingzi and Shuanglongtan landslides) and a large catchment of the Jinsha River, China. By comparing our results with the ones using purely optical imagery, the precision of the Jinpingzi landslide is improved by 14.12%, and the recall and precision of the Shuanglongtan landslide are improved by 3.1% and 3.6%, respectively, and the recall for the large catchment is improved by 9.95%. Our method can improve delineation of potential landslides significantly, which is crucial for landslide early warning and prevention.

Comparative Study on Potential Landslide Identification with ALOS-2 and Sentinel-1A Data in Heavy Forest Reach, Upstream of the Jinsha River

Many SAR satellites such as the ALOS-2 satellite and Sentinel-1A satellite can be used in Interferometric Synthetic Aperture Radar (InSAR) to identify landslides. As their wavelengths are different, they can perform differently in the same area. In this study, we selected the alpine canyon heavy forest area of the Baishugong–Shangjiangxiang section of the Jinsha River with a strong uplift of faults and folds as the study area. The Small Baseline Subset (SBAS)–InSAR was used for landslide identification to compare the reliability and applicability of L-band ALOS-2 data and C-band Sentinel-1A data. In total, 13 potential landslides were identified, of which 12 potential landslides were identified by ALOS-2 data, two landslides were identified by Sentinel-1A data, and the Kongzhigong (KZG) landslide was identified by both datasets. Then, the field investigation was used to verify the identification results and analyze the genetic mechanism of four typical landslides. Both the Duila (DL) and KZG landslides are bedding slip, while the Jirenhe (JRH) and Maopo (MP) landslides are creep–pull failure. Then, the difference between ALOS-2 and Sentinel-1A data on KZG landslide was compared. A total of 35,961 deformation points on the KZG landslide were obtained using ALOS-2 data, which are relatively dense. Meanwhile, a total of 7715 deformation points were obtained by Sentinel-1A data, which are relatively scattered and seriously lacking, especially in areas with dense vegetation coverage. Comparing the advantages of ALOS-2 and Sentinel-1A data and the identification results of potential landslides, the reliability and applicability of ALOS-2 data in the identification of potential landslides in areas with dense vegetation cover and complex geological conditions were confirmed from the aspects of vegetation cover, topography, field investigation, and comparative analysis of typical landslides.

Deep learning-based landslide susceptibility mapping

Landslides are considered as one of the most devastating natural hazards in Iran, causing extensive damage and loss of life. Landslide susceptibility maps for landslide prone areas can be used to plan for and mitigate the consequences of catastrophic landsliding events. Here, we developed a deep convolutional neural network (CNN–DNN) for mapping landslide susceptibility, and evaluated it on the Isfahan province, Iran, which has not previously been assessed on such a scale. The proposed model was trained and validated using training (80%) and testing (20%) datasets, each containing relevant data on historical landslides, field records and remote sensing images, and a range of geomorphological, geological, environmental and human activity factors as covariates. The CNN–DNN model prediction accuracy was tested using a wide range of statistics from the confusion matrix and error indices from the receiver operating characteristic (ROC) curve. The CNN–DNN model was evaluated comprehensively by comparing it to several state-of-the-art benchmark machine learning techniques including the support vector machine (SVM), logistic regression (LR), Gaussian naïve Bayes (GNB), multilayer perceptron (MLP), Bernoulli Naïve Bayes (BNB) and decision tree (DT) classifiers. The CNN–DNN model for landslide susceptibility mapping was found to predict more accurately than the benchmark algorithms, with an AUC = 90.9%, IRs = 84.8%, MSE = 0.17, RMSE = 0.40, and MAPE = 0.42. The map provided by the CNN–DNN clearly revealed a high-susceptibility area in the west and southwest, related to the main Zagros trend in the province. These findings can be of great utility for landslide risk management and land use planning in the Isfahan province.

A Novel Model for Landslide Displacement Prediction Based on EDR Selection and Multi-Swarm Intelligence Optimization Algorithm

With the widespread application of machine learning methods, the continuous improvement of forecast accuracy has become an important task, which is especially crucial for landslide displacement predictions. This study aimed to propose a novel prediction model to improve accuracy in landslide prediction, based on the combination of multiple new algorithms. The proposed new method includes three parts: data preparation, multi-swarm intelligence (MSI) optimization, and displacement prediction. In the data preparation, the complete ensemble empirical mode decomposition (CEEMD) is adopted to separate the trend and periodic displacements from the observed cumulative landslide displacement. The frequency component and residual component of reconstructed inducing factors that related to landslide movements are also extracted by the CEEMD and t-test, and then picked out with edit distance on real sequence (EDR) as input variables for the support vector regression (SVR) model. MSI optimization algorithms are used to optimize the SVR model in the MSI optimization; thus, six predictions models can be obtained that can be used in the displacement prediction part. Finally, the trend and periodic displacements are predicted by six optimized SVR models, respectively. The trend displacement and periodic displacement with the highest prediction accuracy are added and regarded as the final prediction result. The case study of the Shiliushubao landslide shows that the prediction results match the observed data well with an improvement in the aspect of average relative error, which indicates that the proposed model can predict landslide displacements with high precision, even when the displacements are characterized by stepped curves that under the influence of multiple time-varying factors.

Evaluation of Conditioning Factors of Slope Instability and Continuous Change Maps in the Generation of Landslide Inventory Maps Using Machine Learning (ML) Algorithms

Landslides are recognized as high-impact natural hazards in different regions around the world; therefore, they are extensively researched by experts. Landslide inventories are essential to identify areas that are likely to be affected in the future, thereby enabling interventions to prevent loss of life. Today, through combined approaches, such as remote sensing and machine learning techniques, it is possible to apply algorithms that use data derived from satellite images to produce landslide inventories. This work presents the performance of five machine learning methods—k-nearest neighbor (KNN), stochastic gradient descendent (SGD), support vector machine radial basis function (SVM RBF Kernel), support vector machine (SVM linear kernel), and AdaBoost—in landslide detection in a zone of the state of Guerrero in southern Mexico, using continuous change maps and primary landslide factors, such as slope angle, terrain orientation (aspect), and lithology, as inputs. The models were trained with 2/3 of ground truth samples of 671 slidden/non-slidden polygons. The obtained inventory maps were evaluated with the remaining 1/3 of ground truth samples by generating a confusion matrix and applying the Kappa concordance coefficient, accuracy, precision, recall, and F1 score as evaluation metrics, as well as omission and commission errors. According to the results, the AdaBoost classifier reached greater spatial and statistical coherence than the other implemented methods. The best input layer combination for detection was the continuous change maps obtained by the linear regression and image differencing detection methods, together with the slope angle, aspect, and lithology conditioning factors.

Quantitative Assessment of Landslide Risk Based on Susceptibility Mapping Using Random Forest and GeoDetector

This study aims to evaluate risk and discover the distribution law for landslides, so as to enrich landslide prevention theory and method. It first selected Fengjie County in the Three Gorges Reservoir Area as the study area. The work involved developing a landslide risk map using hazard and vulnerability maps utilizing landslide dataset from 2001 to 2016. The landslide dataset was built from historical records, satellite images and extensive field surveys. Firstly, under four primary conditioning factors (i.e., topographic factors, geological factors, meteorological and hydrological factors and vegetation factors), 19 dominant factors were selected from 25 secondary conditioning factors based on the GeoDetector to form an evaluation factor library for the LSM. Subsequently, the random forest model (RF) was used to analyze landslide susceptibility. Then, the landslide hazard map was generated based on the landslide susceptibility mapping (LSM) for the study region. Thereafter, landslide vulnerability assessment was conducted using key elements (economic, material, community) and the weights were provided based on expert judgment. Finally, when risk equals vulnerability multiplied by hazard, the region was categorized as very low, low, medium, high and very high risk level. The results showed that most landslides distribute on both sides of the reservoir bank and the primary and secondary tributaries in the study area, which showed a spatial distribution pattern of more north than south. Elevation, lithology and groundwater type are the main factors affecting landslides. Fengjie County landslide risk level is mostly low (accounting for 73.71% of the study area), but a small part is high and very high risk level (accounting for 2.5%). The overall risk level shows the spatial distribution characteristics of high risk in the central and eastern urban areas and low risk in the southern and northern high-altitude areas. Secondly, it is necessary to strictly control the key risk areas, and carry out prevention and control zoning management according to local conditions. The study is conducted for a specific region but can be extended to other areas around the investigated area. The developed landslide risk map can be considered by relevant government officials for the smooth implementation of management at the regional scale.

Unraveling the drivers of intensified landslide regimes in Western Ghats, India

The Western Ghats (WG) mountain range in the Indian sub-continent is a biodiversity hotspot, now faces a severe threat to the valley population and ecosystem because of changing rainfall patterns and land-use changes. Here, we use the 2018-2019 landslide inventory data together with various geo-environmental factors and show that the landslide activity in the WG region is amplified by anthropogenic disturbances. We applied a generalized feature selection algorithm and a random forest susceptibility model to demonstrate the major topographic controls of landslides and the risk associated with them in the WG region. Our results show that road cutting and slopes modified to plantations are the strongest environmental variable (50% - 73% within 300 m buffer distance) related to the landslide patterns, whereas short-duration intense precipitation in the high elevated terrain, profile concavity, and stream power contributed to the initiation of landslides. The susceptibility models made for the present, and Global Climate Models (GCM) under the representative concentration pathway (RCP) 8.5 scenario predicts the vulnerable nature of WG for future climate extremes. Our results highlight the impacts of Anthropocene hazards and sensitivity of the WG ecosystem, and a greater focus therefore should be placed to reduce the vulnerability and increase preparedness for future events.

Comparison of Machine Learning Methods for Potential Active Landslide Hazards Identification with Multi-Source Data

The early identification of potential landslide hazards is of great practical significance for disaster early warning and prevention. The study used different machine learning methods to identify potential active landslides along a 15 km buffer zone on both sides of Jinsha River (Panzhihua-Huize section), China. The morphology and texture features of landslides were characterized with InSAR deformation monitoring data and high-resolution optical remote sensing data, combined with 17 landslide influencing factors. In the study area, 83 deformation accumulation areas of potential landslide hazards and 54 deformation accumulation areas of non-potential landslide hazards were identified through spatial overlay analysis with 64 potential active landslides, which have been confirmed by field verification. The Naive Bayes (NB), Decision Tree (DT), Support Vector Machine (SVM) and Random Forest (RF) algorithms were trained and tested through attribute selection and parameter optimization. Among the 17 landslide influencing factors, Drainage Density, NDVI, Slope and Weathering Degree play an indispensable role in the machine learning and recognition of landslide hazards in our study area, while other influencing factors play a certain role in different algorithms. A multi-index (Precision, Recall, F1) comparison shows that the SVM (0.867, 0.829, 0.816) has better recognition precision skill for small-scale unbalanced landslide deformation datasets, followed by RF (0.765, 0.756, 0.741), DT (0.755, 0.756, 0.748) and NB (0.659, 0.659, 0.659). Different from the previous study on landslide susceptibility and hazard mapping based on machine learning, this study focuses on how to find out the potential active landslide points more accurately, rather than evaluating the landslide susceptibility of specific areas to tell us which areas are more sensitive to landslides. This study verified the feasibility of early identification of landslide hazards by using different machine learning methods combined with deformation information and multi-source landslide influencing factors rather than by relying on human–computer interaction. This study shows that the efficiency of potential hazard identification can be increased while reducing the subjective bias caused by relying only on human experts.

Uncertainties Analysis of Collapse Susceptibility Prediction Based on Remote Sensing and GIS: Influences of Different Data-Based Models and Connections between Collapses and Environmental Factors

To study the uncertainties of a collapse susceptibility prediction (CSP) under the coupled conditions of different data-based models and different connection methods between collapses and environmental factors, An’yuan County in China with 108 collapses is used as the study case, and 11 environmental factors are acquired by data analysis of Landsat TM 8 and high-resolution aerial images, using a hydrological and topographical spatial analysis of Digital Elevation Modeling in ArcGIS 10.2 software. Accordingly, 20 coupled conditions are proposed for CSP with five different connection methods (Probability Statistics (PSs), Frequency Ratio (FR), Information Value (IV), Index of Entropy (IOE) and Weight of Evidence (WOE)) and four data-based models (Analytic Hierarchy Process (AHP), Multiple Linear Regression (MLR), C5.0 Decision Tree (C5.0 DT) and Random Forest (RF)). Finally, the CSP uncertainties are assessed using the area under receiver operation curve (AUC), mean value, standard deviation and significance test, respectively. Results show that: (1) the WOE-based models have the highest AUC accuracy, lowest mean values and average rank, and a relatively large standard deviation; the mean values and average rank of all the FR-, IV- and IOE-based models are relatively large with low standard deviations; meanwhile, the AUC accuracies of FR-, IV- and IOE-based models are consistent but higher than those of the PS-based model. Hence, the WOE exhibits a greater spatial correlation performance than the other four methods. (2) Among all the data-based models, the RF model has the highest AUC accuracy, lowest mean value and mean rank, and a relatively large standard deviation. The CSP performance of the RF model is followed by the C5.0 DT, MLR and AHP models, respectively. (3) Under the coupled conditions, the WOE-RF model has the highest AUC accuracy, a relatively low mean value and average rank, and a high standard deviation. The PS-AHP model is opposite to the WOE-RF model. (4) In addition, the coupled models show slightly better CSP performances than those of the single data-based models not considering connect methods. The CSP performance of the other models falls somewhere in between. It is concluded that the WOE-RF is the most appropriate coupled condition for CSP than the other models.