Landslide Susceptibility Research Articles

Benchmarking data handling strategies for landslide susceptibility modeling using random forest workflows

Machine learning (ML) algorithms are frequently used in landslide susceptibility modeling. Different data handling strategies may generate variations in landslide susceptibility modeling, even when using the same ML algorithm. This research aims to compare the combinations of inventory data handling, cross validation (CV), and hyperparameter tuning strategies to generate landslide susceptibility maps. The results are expected to provide a general strategy for landslide susceptibility modeling using ML techniques. The authors employed eight landslide inventory data handling scenarios to convert a landslide polygon into a landslide point, i.e., the landslide point is located on the toe (minimum height), on the scarp (maximum height), at the center of the landslide, randomly inside the polygon (1 point), randomly inside the polygon (3 points), randomly inside the polygon (5 points), randomly inside the polygon (10 points), and 15 m grid sampling. Random forest models using CV–nonspatial hyperparameter tuning, spatial CV–spatial hyperparameter tuning, and spatial CV–forward feature selection–no hyperparameter tuning were applied for each data handling strategy. The combination generated 24 random forest ML workflows, which are applied using a complete inventory of 743 landslides triggered by Tropical Cyclone Cempaka (2017) in Pacitan Regency, Indonesia, and 11 landslide controlling factors. The results show that grid sampling with spatial CV and spatial hyperparameter tuning is favorable because the strategy can minimize overfitting, generate a relatively high-performance predictive model, and reduce the appearance of susceptibility artifacts in the landslide area. Careful data inventory handling, CV, and hyperparameter tuning strategies should be considered in landslide susceptibility modeling to increase the applicability of landslide susceptibility maps in practical application.

Landslide susceptibility in Phuoc Son, Quang Nam: A deep learning approach

Landslide susceptibility in Phuoc Son, Quang Nam: A deep learning approach

Exploiting the land use to predict shallow landslide susceptibility: A probabilistic implementation of LAPSUS-LS

Due to the significant role of land use on the occurrence of rainfall-induced shallow landslides, this factor is commonly employed as a landslide susceptibility predictor. However, the land use classification is oftentimes very broad, neglecting the proven mechanical and hydrogeological role of the land management on slope stability. Given the necessity of including spatially distributed and management-specific inputs, the process-based landscape evolution model LAPSUS-LS was chosen and adapted to achieve a probabilistic approach which takes into account land management as an input by adopting management-specific values of root cohesion. The model was applied to four test sites in the Oltrepò Pavese (Italy), where different vineyard management techniques play a significant role in triggering landslides. The results for the four test areas had, cumulatively, an Area Under the Roc curve greater than 0.73, with false negative cells being < 1 % of the total for all simulations. In the model’s application, land use practices characterised by higher root cohesion proved to benefit slope stability, whereas tilled vineyards, shrublands and abandoned vineyards were more prone to the formation of shallow landslides. In addition, we found that the inclusion of management-specific input parameters produced more accurate outputs and that in catchments characterised by average slope angles lower than 15°, varying the vineyard management, did not appear to affect the landslide susceptibility. Due to the model’s high dependency on the land use and its ability to include land management, it can take into account the spatial variability of input values such as the root cohesion. Additionally, it can be applied i) to manage current conditions, ii) to explore future land use change, iii) to study less invasive yet beneficial land use management change scenarios and iv) provide farmers of at-risk areas insight on how to improve slope stability.

Construction of landslide warning by combining rainfall threshold and landslide susceptibility in the gully region of the Loess Plateau: A case of Lanzhou City, China

Regional forecasting of the landslide rainfall threshold is complex, and the empirical prediction of the rainfall threshold relies on historical statistical data. To obtain the intensity I ∼ duration (D) rainfall threshold, Lanzhou City was selected as the research object, and a field investigation, 3S technology, mathematical statistics and a Python language environment were employed to develop a set of automatic rainfall threshold calculation programs. Factor elimination, the information value method, weight of evidence, the frequency ratio, the density method and receiver operating characteristic (ROC) curve were applied to obtain landslide susceptibility, and the I ∼ D rainfall threshold (temporal probability) and landslide susceptibility (spatial probability) were combined to construct regional landslide rainfall threshold warning levels (red, orange, yellow, and blue). The results showed the rainfall duration of historical landslides ranged from 0.99 to 216 h, thereinto, it was accounted for 92.78 % with that of between 6 and 96 h, and the rainfall intensity was between 0.11 and 23.6 mm/h, and the α and β of the rainfall threshold curve was gradually increased with an increase in the rainfall threshold probability. Additionally, when the rainfall condition probability was the same, landslides occurrence was more frequent in Chenguan District, where there are large areas of bare land, and slope and annual rainfall are less than 10 °and 300 mm, respectively. Fifteen, nine, and six factors were used to obtain landslide susceptibility, and the accuracy of the ROC ranged from 0.78 to 0.80 under different methods and different factor combinations. The spatial distribution of landslide susceptibility was very similar. There were 35, 12, 25, and 108 landslides associated with red, orange, yellow, and blue levels, respectively, under the spatial–temporal probability, but 30, 26, 54, and 70, respectively, under temporal probability. Under the different conditions, there were 89, 69, and 22 landslide points with unchanged, increased, and decreased warning levels, respectively. The ability and accuracy of rainfall threshold warning and prediction were effectively improved by integrating the temporal and spatial probabilities of the rainfall threshold.

Spatial Prediction of Landslide Susceptibility using Various Machine Learning Based Binary Classification Methods

ABSTRACT This study compares the performance of various machine learning models for predicting landslide susceptibility using a geospatial dataset from the Lai Chau province, Vietnam. The dataset consisted of 850 landslide locations and ten influencing factors. Eight models, including Forest by Penalizing Attributes (FPA), Bagging-based FPA (BFPA), Artificial Neural Network (ANN), Logistic Regression (LR), Support Vector Machine (SVM), Multilayer Perceptron (MLP), Bayesian Network (BN), and Naïve Bayes (NB), were evaluated based on different evaluation metrics. The results revealed distinct variations in the performance of the models across the evaluation metrics. Based on the overall rankings, the ensemble BFPA model with sensitivity=90%, specificity= 95.98%, accuracy=92.86%, Kappa=0.857, and area under the curve=0.98 demonstrated the highest capability in predicting landslide susceptibility. It was followed by BN, FPA, MLP, ANN, SVM, LR, and NB. These findings suggest that the BFPA model outperformed other models in terms of its ability to accurately identify potential landslide-prone areas in the study region. This study provides valuable insights into the comparative analysis of machine learning models for landslide susceptibility prediction. Furthermore, it supports the effectiveness of ensemble models for landslide susceptibility prediction, which can inform decision-makers, land-use planners, and disaster management agencies in making informed decisions regarding potential landslide hazards and implementing effective risk mitigation strategies in Vietnam. Continued research in this area will enhance our understanding of machine learning techniques and their application in mitigating the impact of landslides on society and the environment.

Landslide susceptibility assessment along highways (SH-12 and NH-717A) in Darjeeling Himalayas

Landslides are the utmost damaging natural along with anthropogenic hazards which cause huge damage to assets and human lives in hilly environments all over the world. The key aim of this study is to explore landslide susceptibility (LS) maps along highways (SH-12 and NH-717A) in the Darjeeling Himalayas, India. Four data-driven bivariate statistical models (BSMs) were used to develop LS maps based on geospatial techniques. A total of 90 (100%) polygons with different size of landslide locations were identified to develop landslide inventory/distribution (LI/LD) map by incorporating Google Earth and satellite imageries, and filed investigation through the global positioning system (GPS). The LI/LD map is distributed into testing 30% (27 landslide locations), and training 70% (63 landslide locations) of the models. Eighteen landslide causative factors (LCFs) were selected to develop LS maps. Multicollinearity analysis showed that there is no collinearity problem in the LCFs. The receiver operating characteristics (ROC) curve and frequency ratio (FR) value techniques were applied for validation. The relative importance of each LCF for each model has been measured using Jack-knife test. The LS maps were divided into six landslide susceptibility zones (LSZs). The ROC curve showed the accuracy of the LSZ maps is 93.50% for FR, 85.90% for evidential belief functions (EBFs), 81.80% for fuzzy membership (FM) using FR, and 77.20% for FM using cosine amplitude (CA) approaches respectively. Among all models, the FR model showed the highest percentage of accuracy for LS assessment and prediction. The FR values of six LSZs rise from very low to very high LSZs which indicate a positive correlation between LSZs and FR values. The results of Jack-knife test showed that terrain ruggedness index (TRI) factor has the highest relative importance for all models in LSZ mappings. The study findings will help to the planner, and policymakers for spatial planning, implement slope management strategies, land use management in mountainous environments.

Evaluating landslide susceptibility: the impact of resolution and hybrid integration approaches

The present study investigates the effectiveness of various landslide susceptibility machine learning (ML) models at multiple spatial resolutions. Using various conditioning factors, including topography, hydrology, and human influences, the study analyzed the predictive power of single, integrated, and comparative ML models. Alternating Decision Tree (ADT), Forest by Penalizing Attributes (FPA), and Random Forest (RAF), and integrated approaches such as Rotation Forest (RF) and Random Subspace (RS) that were based on ADT and FPA were utilized for the study. All models were trained and validated using a ten-fold cross-validation technique and evaluated by various statistical metrics, across resolutions from 12.5 m to 200 m. Shenmu City in China was chosen as an ideal test site to evaluate the developed methodology. The study reveals that finer spatial resolutions significantly enhance the accuracy of landslide predictions and that integrated models have superior performance over single models. The Frequency Ratio method identified elevation, slope, and hydrological factors as the key predictors concerning landslide occurrences. According to the results the RS-ADT model at a 12.5 m resolution achieved the highest evaluation accuracy (ROC = 0.907). The research highlights the possibility of combining multiple modeling techniques and high-resolution data to improve the predictive accuracy of landslide susceptibility models.

A Novel Strategy Coupling Optimised Sampling with Heterogeneous Ensemble Machine-Learning to Predict Landslide Susceptibility

The accuracy of data-driven landslide susceptibility prediction depends heavily on the quality of non-landslide samples and the selection of machine-learning algorithms. Current methods rely on artificial prior knowledge to obtain negative samples from landslide-free regions or outside the landslide buffer zones randomly and quickly but often ignore the reliability of non-landslide samples, which will pose a serious risk of including potential landslides and lead to erroneous outcomes in training data. Furthermore, diverse machine-learning models exhibit distinct classification capabilities, and applying a single model can readily result in over-fitting of the dataset and introduce potential uncertainties in predictions. To address these problems, taking Chenxi County, a hilly and mountainous area in southern China, as an example, this research proposes a strategy-coupling optimised sampling with heterogeneous ensemble machine learning to enhance the accuracy of landslide susceptibility prediction. Initially, 21 landslide impact factors were derived from five aspects: geology, hydrology, topography, meteorology, human activities, and geographical environment. Then, these factors were screened through a correlation analysis and collinearity diagnosis. Afterwards, an optimised sampling (OS) method was utilised to select negative samples by fusing the reliability of non-landslide samples and certainty factor values on the basis of the environmental similarity and statistical model. Subsequently, the adopted non-landslide samples and historical landslides were combined to create machine-learning datasets. Finally, baseline models (support vector machine, random forest, and back propagation neural network) and the stacking ensemble model were employed to predict susceptibility. The findings indicated that the OS method, considering the reliability of non-landslide samples, achieved higher-quality negative samples than currently widely used sampling methods. The stacking ensemble machine-learning model outperformed those three baseline models. Notably, the accuracy of the hybrid OS–Stacking model is most promising, up to 97.1%. The integrated strategy significantly improves the prediction of landslide susceptibility and makes it reliable and effective for assessing regional geohazard risk.

Machine learning and deep learning-based landslide susceptibility mapping using geospatial techniques in Wayanad, Kerala state, India

Landslide susceptibility mapping is vital for disaster management and sustainable land-use planning. This research was conducted in Wayanad, Kerala, India, to identify landslide susceptible zones. The study used large geospatial datasets, such as elevation, slope, aspect, curvature, stream power index, topographic wetness index, land use and land cover, rainfall, flow accumulation, geology, and geomorphology. It is followed by the application of various machine learning and deep learning models such as the support vector machine, artificial neural networks, logistic regression, random forest, gradient boosting machine, recurrent neural networks long short-term memory, and deep neural network models to map the landslide susceptible zones. The model was trained and validated using the landslide inventory map, which contains 298 sites of landslides. The random forest model, with 97 % accuracy, performed best. It is possible to effectively mitigate landslides and plan long-term land use by identifying hazardous zones within the study region.

Geospatial Intelligence for Landslide Susceptibility and Risk Analysis: Insights from NH31A and East Sikkim Himalaya Settlements

Slope instability is a serious concern in the Sikkim Himalayas. The town and numerous road segments along National Highway 31A were ravaged by multiple landslides that occurred in the nearby region. A bivariate statistical method known as frequency ratio (FR), information value (IV), and certainty factor (CF) analysis was employed in this work to examine landslide risk assessment (LRA) and landslide susceptibility zonation (LSZ) maps in the Rorachu watershed. This study represents the first comprehensive analysis of landslide risk in the populated areas of East Sikkim and along NH31A, offering a deeper understanding of the risks involved and contributing to the enhancement of local resilience against landslide hazards. A total of 153 different landslide locations were mapped using Google Earth and GIS software; 30% (46) of these locations were used to validate the models, and 70% of these (107) served as training data for the FR, IV, and CF models. The thirteen landslide causative factors (geology, soil, elevations, slope, curvature, drainage density (DD), road density (RD), rainfall, normalized difference vegetation index (NDVI), land use land cover (LULC), topographic position index (TPI), stream power index (SPI), and topographic wetness index (TWI)) were extracted from a spatial database for LSZ mapping. Landslides were most prevalent on slopes (35° to 50°), heights (2,500–4,100 m), and rainfall (2000–2,500 mm and 3,000–3,300 mm). The area under the curves (AUC) for the FR, IV, and CF models are 0.925 (92.50%), 0.846 (84.60%), and 0.868 (86.20%), respectively. The prediction rates are shown by the AUCs for the FR, IV, and CF models, which are 0.828 (82.8%), 0.750 (%), and 0.836 (83.60%), respectively. According to the landslide risk assessment (LRA), the FR (20.75%), IV (40.91%) and CF (18.78%) models showed high risk on Highway 31A, while the FR (9.05%), IV (38.59%) and CF (20.90%) models showed high risk in densely populated areas. These landslide risk and vulnerability maps can be used to develop land use planning strategies that can save lives and are useful for planners and mitigation measures. Special attention should be paid to urbanization, highway construction, and deforestation.

Machine learning approaches for mapping and predicting landslide-prone areas in São Sebastião (Southeast Brazil)

This study employs machine learning techniques to map and predict landslide-prone areas in São Sebastião, Brazil, a region susceptible to landslides due to its steep terrain and intense rainfall. We compared five algorithms: Random Forest, Gradient Boosting, Support Vector Machine, Artificial Neural Network, and k-Nearest Neighbors, using various environmental factors as inputs. The Gradient Boosting model performed best, achieving an AUC-ROC of 0.963 and an accuracy of 99.6%. Slope degree, soil moisture index, and relief dissection emerged as the most influential factors in predicting landslide susceptibility. Analysis of land use and land cover changes between 1985 and 2021 revealed significant increases in forest cover and urban areas, with implications for landslide risk distribution. The resulting susceptibility map shows predominantly low-risk areas with scattered high-risk zones, providing crucial information for targeted risk management. This research demonstrates the effectiveness of machine learning in landslide susceptibility mapping and offers valuable insights for disaster risk reduction and urban planning in coastal mountainous regions.

Landslide risk assessment in mining areas using hybrid machine learning methods under fuzzy environment

Landslide risk assessment in coal mining economic cities faces challenges due to inadequate knowledge of indicator systems and modeling techniques, despite its crucial significance in geological disaster management. This study addresses this gap by proposing a novel framework encompassing data preparation, susceptibility analysis, consequence analysis, and final risk assessment. The primary objective is to reduce the complexity and inherent uncertainties associated with landslide risk assessment in these specific urban environments. First, two comprehensive databases for the indication system were established. One database focuses on landslide susceptibility, considering the region’s mining disturbances and disaster-prone environments. The other evaluates the effect of consequences while accounting for the pertinent risk variables. Second, a hybrid machine learning model called Geo-FR-SVM was developed by combining the GeoDetector, Frequency Ratio method, and Support Vector Machine model to improve landslide susceptibility mapping. Integrating the susceptibility map with the consequence map created using fuzzy set theory and fuzzy analytical hierarchy process methodologies resulted in the final landslide risk map. Our analysis revealed that elevation, distance to rivers and roads, groundwater level, and coal mining subsidence were the key factors influencing landslide occurrence in the Xishan mine area. The majority of the study area exhibits low landslide risk. However, higher consequence zones were primarily concentrated in highly populated residential regions and coal mine industrial zones in the east-central and western urban areas. Additionally, they are mainly distributed along the roads in the center region. With comparatively large percentages of landslide high-risk and extremely high-risk regions (4.596%, 3.855%, 3.625%, and 3.188%, respectively) at Tunlan Mine, Xiqu Mine, Zhenchengdi Mine, and Baijiazhuang Mine, these findings highlight the need for these mines to prioritize potential landslide identification and risk mitigation strategies. This framework offers broad applicability to other coal mining economic cities, serving as a valuable foundation for landslide risk management decision-making.

GIS-based statistical and limit equilibrium models in the assessment of slope stability and landslide susceptibility: the case study of the Aomar Miocene basin, Bouira, Algeria

GIS-based statistical and limit equilibrium models in the assessment of slope stability and landslide susceptibility: the case study of the Aomar Miocene basin, Bouira, Algeria

Comparing the effectiveness of landslide susceptibility mapping by using the frequency ratio and hybrid MCDM models

The present study involved the landslide susceptibility mapping of Indian state of Meghalaya by using the frequency ratio (FR) and Analytical Hierarchy Process (AHP)- VlseKriterijuska Optimizacija I Komoromisno Resenje (VIKOR) (AHP-VIKOR), and the outcomes were validated and compared. Therefore, fourteen landslide conditioning factors were selected after a multi-collinearity test by tolerance and variance Inflation Factor. Based on 70 % of the training data, the relative frequency and predictive rate value was used to generate final output. In opposite, the overall ranking was used to generate landslide susceptibility map (LSM) for the AHP-VIKOR model. Subsequently, the generated LSM was classified into five hierarchical zones from very low to very high based on the natural breaking classification. The area under the curve of receiver operating characteristics from 30 % of validation landslides shown more predictive accuracy of AHP-VIKOR model than FR model under similar geo-environmental condition. The spatial distribution of landslide susceptibility zones shows that 37 % of the region under the threat of having a high landslip risk. Landslides are primarily found in barren Cretaceous sedimentary rocks, with increased risk near roadsides. This is due to slope cutting for roads, making hilly terrain unstable. Areas with more annual rainfall face greater landslide threats. Therefore, the outcome of the study may contain the useful information that helps the local administration to recognize vulnerability of landslides as well as to take up the hazard mitigative measures.

Co-seismic landslide susceptibility mapping for the Luding earthquake area based on heterogeneous ensemble machine learning models

ABSTRACT Considering the complexity of landslide susceptibility mapping (LSM) at the municipal-scale, it is difficult for the assumption space of a single machine learning (ML) algorithm to fully and accurately reflect the intrinsic correlation between the landslide influencing factors and the actual catastrophic events under all scenarios. This study addresses this by integrating three heterogeneous ensemble learning (HEL) algorithms(Bagging, Blending, and Stacking) with four ML models (Gradient Boosting Decision Tree (GBDT), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), and Backpropagation-Neural Network (BPNN)), to construct 27 HEL LSM models, and to realize optimization and selection of the models and algorithms through the comparative analysis of the results of the LSM and the accuracy of the prediction of the individual models. The final results indicate that the Bagging and Blending HEL algorithms fail to enhance model prediction accuracy effectively. In contrast, the Stacking HEL algorithm significantly improves model prediction accuracy (with an accuracy increase of over 1.4% and AUC values above 0.95). The Stacking (GBDT-XGBoost) model had the highest prediction accuracy (AUC = 0.964) and thus was selected as the optimal LSM model. This study's modeling and mapping results are intended to serve as a reference for geohazard prevention and sustainability studies in related cities.

A QGIS framework for physically-based probabilistic modelling of landslide susceptibility: QGIS-FORM

A QGIS framework for physically-based probabilistic modelling of landslide susceptibility: QGIS-FORM

Development of shallow landslide susceptibility maps incorporating relative spacing index for forest management

Development of shallow landslide susceptibility maps incorporating relative spacing index for forest management

Landslides: A Review from the Southern Western Ghats of India

ABSTRACT Landslides are the most unpredictable catastrophic events in mountainous and hilly regions. South Western Ghats one among of the regions in India that have experienced recurring landslides due to increase in pre-and post-monsoon rains. Landslides often cause damage to buildings, roads, natural vegetation, even result in the loss of human lives in South Western Ghats with indirect impact on the socioeconomic status of the country. It is crucial to depict the current trend technologies and their performance in landslide studies, offering valuable insights into mitigation strategies. This article highlights on prevalent techniques for mapping landslide zones, along with major conditional factors and assess the performance of these techniques in landslide susceptibility mapping of the South Western Ghats. Landslide studies of the South Western Ghats reveal 20% of the research work with the Landsat 8 OLI for mapping landslides, 13.3% of the studies using the higher spatial resolutions 2.5m and 10m of Cartosat DEM, the rest of the studies were based on the spatial resolution of 12.5m and 30m from ALOS PALSAR, ASTER and SRTM DEM. About 26% landslide studies have employed quantitative approaches primarily the frequency ratio of bivariate. Subsequently machine learning techniques such as random forest and support vector machine have gained prominence, accounting for 33% landslide studies. The literature depicts precipitation as a significant triggering factor and slope as the major conditional factor in the occurrence of landslides within the South Western Ghats. Prioritized densely populated areas in landslide studies have increased attention to smaller geographical areas and forest regions in the South Western Ghats to protect both inhabitants and ecosystems.

Fuzzy Logic-Based Landslide Susceptibility Mapping in Earthquake-Prone Areas: A Case Study of the Mila Basin, Algeria

Abstract —This research focuses on analyzing landslides triggered by a moderate earthquake (Mw = 4.9) in the northeastern region of the Mila province, which resulted in significant damage and economic losses in the El Kherba district and Grarem Gouga city. Through an extensive field-based investigation, a comprehensive inventory of landslides has been compiled. To assess the susceptibility to landslides triggered by seismic activity, a GIS-based fuzzy logic model was employed. The model incorporates various input factors, such as lithology, slope angle, normalized difference vegetation index (NDVI), distance from rivers and roads, precipitation, and seismic hazard, which is shown on a map. The study compares the performance of different fuzzy operators and gamma values and determines that using fuzzy gamma operators with a gamma value of 0.8 yields a satisfactory consistency with the distribution of landslides. Moreover, incorporating the map of seismic hazard as a causative factor enhances the accuracy of landslide susceptibility mapping. This study underscores the utility of the fuzzy logic model in disaster management and the planning of development activities.

Evaluation of statistical modeling (SM) approaches for landslide susceptibility mapping: geospatial insights for Bhutan

Evaluation of statistical modeling (SM) approaches for landslide susceptibility mapping: geospatial insights for Bhutan