Landslide Susceptibility Zonation Research Articles

Comprehensive landslide prediction mapping using bivariate statistical models of Mizoram state of Northeast India

ABSTRACT Landslides in the state of Mizoram result in damage to life and properties annually. The study focuses on landslide susceptibility zones by frequency ratio (FR), evidential belief function (EBF) and index of entropy (IOE) models. A total of 1,486 landslide points were used to build a relationship between 16 factors and landslide occurrences. The results reveal 14.44%, 19.64% and 3.55% of the area as very high susceptible zones in FR, EBF and IOE models, respectively. The AUC results support the adoption of the EBF model in land use planning and decision-making processes to enhance natural resource management and mitigate landslide risks in Mizoram.

Game-theoretic optimization of landslide susceptibility mapping: a comparative study between Bayesian-optimized basic neural network and new generation neural network models.

Landslide susceptibility mapping is essential for reducing the risk of landslides and ensuring the safety of people and infrastructure in landslide-prone areas. However, little research has been done on the development of well-optimized Elman neural networks (ENN), deep neural networks (DNN), and artificial neural networks (ANN) for robust landslide susceptibility mapping (LSM). Additionally, there is a research gap regarding the use of Bayesian optimization and the derivation of SHapley Additive exPlanations (SHAP) values from optimized models. Therefore, this study aims to optimize DNN, ENN, and ANN models using Bayesian optimization for landslide susceptibility mapping and derive SHAP values from these optimized models. The LSM models have been validated using the receiver operating characteristics curve, confusion matrix, and other twelve error matrices. The study used six machine learning-based feature selection techniques to identify the most important variables for predicting landslide susceptibility. The decision tree, random forest, and bagging feature selection models showed that slope, elevation, DFR, annual rainfall, LD, DD, RD, and LULC are influential variables, while geology and soil texture have less influence. The DNN model outperformed the other two models, covering 7839.54 km2 under the very low landslide susceptibility zone and 3613.44 km2 under the very high landslide susceptibility zone. The DNN model is better suited for generating landslide susceptibility maps, as it can classify areas with higher accuracy. The model identified several key factors that contribute to the initiation of landslides, including high elevation, built-up and agricultural land use, less vegetation, aspect (north and northwest), soil depth less than 140 cm, high rainfall, high lineament density, and a low distance from roads. The study's findings can help stakeholders make informed decisions to reduce the risk of landslides and ensure the safety of people and infrastructure in landslide-prone areas.

Landslide susceptibility modelling in the Doda Kishtwar ramban (DKR) region of Jammu and Kashmir using remote sensing and geographic information system

The present study represents a significant understanding, employing cutting-edge Remote Sensing(RS) and Geographic Information System (GIS) technologies by conducting a comprehensive landslide susceptibility modelling in the Doda, Kishtwar, and Ramban (DKR) districts of the Union Territory of the Jammu and Kashmir in the NW Himalaya. The primary objective of this study is to determine the areas that are at high risk of landslides and accordingly classified these risk zones into five distinct categories, ranging from very high (VH) to very low (VL) susceptibility. To achieve this, a multifaceted approach, involving the utilization of various terrain thematic layers in the GIS environment was applied. The preferred framework employed in this study is the Weighted Overlay (WO) analysis. The final outcome of the study is the final landslide susceptibility map which is a significant and valuable resource for a wide range of stakeholders, including decision-makers, land managers, and local communities. It equips these stakeholders with the information and insights required to adopt proactive strategies in response to the identified landslide susceptibility zones.

Landslide Susceptibility Zoning: Integrating Multiple Intelligent Models with SHAP Analysis

In this study, we aim to delineate landslide susceptibility zones within Dien Bien province, Vietnam, leveraging the capabilities of various machine learning models including Light Gradient Boosting Machine (LGBM), K-Nearest Neighbors (KNN), and Gradient Boosting (GB). Harnessing a dataset comprising 665 data points and encompassing 14 influential factors such as slope, aspect, curvature, elevation, geological composition, Normalized Difference Vegetation Index (NDVI), and proximity to geological features like faults, rivers, and roads, a comprehensive database for landslide modeling was constructed. The analysis entailed rigorous evaluation and comparison of model accuracy employing established statistical metrics, notably Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC). The findings underscore the efficacy of the Light Gradient Boosting Machine model, exhibiting superior performance with an AUC score of 0.85, surpassing both the Gradient Boosting model (AUC = 0.81) and the K-Nearest Neighbors model (AUC = 0.79). Notably, the Light Gradient Boosting Machine model emerges as a promising tool for precise landslide prediction within the study area, offering significant potential for the creation of a robust landslide susceptibility map. The resulting spatial forecast map for Dien Bien province holds considerable utility for informing land use planning initiatives aimed at mitigating the impact of landslide disasters in the region. Moreover, the application of SHAP (Shapley Additive explanation) values to quantify the contribution of each factor to landslide susceptibility prediction, offering novel insights into model interpretation and feature importance. The resulting spatial forecast map holds significant implications for land use planning and disaster mitigation efforts in Dien Bien province, showcasing the potential of advanced machine learning techniques in enhancing landslide risk management strategies.

Refined landslide susceptibility mapping in township area using ensemble machine learning method under dataset replenishment strategy

Landslide susceptibility mapping (LSM) warrants considerable attention as a prerequisite for risk assessment and prevention in response to the increasing global occurrences of landslides. However, refined LSM at the township scale still faces the problem of the limited number of landslides and the oversight of relevant factors associated with different landslide types. This work is dedicated to propose a novel dataset replenishment strategy designed to address landslide sample scarcity, and to identify dominant factors in high landslide susceptibility zones to refine LSM at township scale for decision makers. Herein, the Wuling County in the Three Gorges Reservoir area of China was taken as taken as the research case. Initially, field investigation and calibration of 21 landslides yielded accumulation and rock areas, and the corresponding impact factors were therefore proposed to construct input features for machine learning facilitated by factor diagnosis using Pearson correlation coefficient (PCC) and multicollinearity test. Subsequently, a training and validation dataset split of 8: 2 was randomly generated, enabling the application of ensemble machine learning model for LSM prediction. The study culminated in the feasibility of dataset replenishment strategy, which was employed to explore uncertainties and identify governing factors in LSM. The results indicate that the bagging-ANN model exhibits superior performance in both accumulation and rock areas, with Areas Under Curve (AUC) of 0.987 and 0.994. Furthermore, the dataset replenishment strategy emerges as instrumental in enhancing the reliability of LSM, as evidenced by an 8.97 % reduction in average value (AVG) and a 11.00 % increase in standard deviation (STD). The governing factors for landslides in accumulation and rock areas are designated as river erosion and height, respectively. Together with the excellent performance, this study is expected to provide a promising reference for LSM in other townships worldwide.

Investigating the dynamic nature of landslide susceptibility in the Indian Himalayan region.

Landslide susceptibility zonation (LSZ) mapping is used to delineate areas prone to landslides and is critical for effective landslide hazard management. The existing methodologies for generating such maps tend to neglect the influence of dynamic environmental variables on landslide occurrences, which may lead to obsolete and erroneous estimates of landslide susceptibility (LS) for a concerned area. Although recent studies have started to report the effects of Land Use/ Land Cover (LULC) variation on LSZ mapping, variations in other dynamic variables like rainfall, soil moisture, and evapotranspiration apart from LULC may also influence slope stability in mountainous regions. The present study investigates the impact of variations in these four variables on the LS distribution, of a selected Indian Himalayan region between 2017 and 2021. Random Forest (RF) susceptibility models are utilized for evaluating the LS for the selected years and geospatial technologies are employed for LS change detection. The results indicate up to 19% variations in the spatial extent for some of the zones of the generated LSZ maps. The research findings of this study are crucial since they reveal the impact of dynamic behavior on LS, which has not been previously documented in the literature.

Landslide susceptibility zonation using the analytical hierarchy process (AHP) in the Great Xi’an Region, China

This study aims to delineate landslide susceptibility maps using the Analytical Hierarchy Process (AHP) method for the Great Xi’an Region, China, which is a key planning project for urban construction in Shaanxi Province, China from 2021 to 2035. Multiple data as elevation, slope, aspect, curvature, river density, soil, lithology, and land use have been considered for delineating the landslide susceptibility maps. Spatially thematic layers and distributed maps of all the aforementioned parameters were created in a GIS environment. Determine the relative importance of these thematic layers in the occurrence of landslides in the study area concerning historical landslide data to assign appropriate weights. Landslide sensitivity maps were generated by a weighted combination in a GIS environment after being analyzed by the AHP method. The sensitivity maps were categorized as “very high (11.06%), high (19.41%), moderate (23.03%), low (28.70%), and very low (17.80%)”. Overlay analysis of the test data with the LSM showed that the moderate to very high landslide susceptibility zones were able to contain 82.58% of the historic landslides. The results of the study help determine the landslide-prone areas in the area and provide a reference for subsequent construction. In addition, the analysis of landslide susceptibility in the area contributes to the study of landslides in similar loess sites.

Exploring machine learning and statistical approach techniques for landslide susceptibility mapping in Siwalik Himalayan Region using geospatial technology.

Landslides are a natural threat that poses a severe risk to human life and the environment. In the Kumaon mountains region in Uttarakhand (India), Nainital is among the most vulnerable areas prone to landslides inflicting harm to livelihood and civilization due to frequent landslides. Developing a landslide susceptibility map (LSM) in this Nainital area will help alleviate the probability of landslide occurrence. GIS and statistical-based approaches like the certainty factor (CF), information value (IV), frequency ratio (FR) and logistic regression (LR) are used for the assessment of LSM. The landslide inventories were prepared using topography, satellite imagery, lithology, slope, aspect, curvature, soil, land use and land cover, geomorphology, drainage density and lineament density to construct the geodatabase of the elements affecting landslides. Furthermore, the receiver operating characteristic (ROC) curve was used to check the accuracy of the predicting model. The results for the area under the curves (AUCs) were 87.8% for logistic regression, 87.6% for certainty factor, 87.4% for information value and 84.8% for frequency ratio, which indicates satisfactory accuracy in landslide susceptibility mapping. The present study perfectly combines GIS and statistical approaches for mapping landslide susceptibility zonation. Regional land use planners and natural disaster management will benefit from the proposed framework for landslide susceptibility maps.

Development of landslide susceptibility maps of Tripura, India using GIS and analytical hierarchy process (AHP).

Landslides are one of the most extensive and destructive geological hazards on the globe. Tripura, a northeastern hilly state of India experiences landslides almost every year during monsoon season causing casualties and huge economic losses. Hence, it is required to assess the landslide susceptibility of the area that would support short- and long-term planning and mitigation. The analytic hierarchy process (AHP) integrated with geospatial technology has been adopted for landslide susceptibility mapping in the state. Eight influencing factors such as slope, lithology, drainage density, rainfall, land use land cover, distance from rivers and roads, and soil type were selected to map the landslide susceptibility. Landslide susceptibility index (LSI) was found to vary from 6.205 during monsoon to 1.427 during post-monsoon season. The LSI values were classified into very high, high, moderate, low, and very low susceptibility. Landslide susceptibility maps for three different seasons, namely, pre-monsoon, monsoon, and post-monsoon, were prepared. The study showed that most of the areas of the state come under very low to moderate landslide susceptibility zones. Around 73.2% area of the state is found to be under low landslide-susceptible zones during the pre-monsoon season, around 62% area is prone to landslides with moderate susceptibility during the monsoon season, and 68.5% area comes under landslides with low susceptibility zones during the post-monsoon season. The resultsof this study may be referred to the engineers and planners for the assessment, control, and mitigation of landslides and the development of basic infrastructure in the state.

An AHP GIS-based Methodology for the Stability Assessment of the Djebel El Ouahch Collapsees on the Numidian Flysh Formation in Northeast Algeria's Constantine Region

This study presents a comprehensive investigation into the factors underlying the collapse incident that occurred in the Jebel El Ouahch Tunnel in the Numidian flysch, Constantine region, Northeast Algeria. This methodology focuses on evaluating landslide susceptibility through the application of the Analytic Hierarchy Process (AHP) along the reconfigured path of the collapsed A1 highway tunnel section in the Constantine region of northeastern Algeria. Various influential factors contributing to landslides were analyzed, including lithofacies, slope gradient, slope aspect, elevations, fault density, plan curvature, distance from streams, and distance from roads. Utilizing a Geographic Information System (GIS), these eight causative factors were prepared for assessment. The findings indicate that slope gradient and lithology play pivotal roles as primary controlling factors in landslide susceptibility. The model exhibited a commendable success rate of 93% in predicting landslide susceptibility, as demonstrated by the area under the curve (AUC) plot generated from the landslides susceptibility map. Most of the new road falls within the highly susceptible area to landslides. This validated model can serve as an effective tool for mapping landslide susceptibility zones along the newly established road path following the collapsed tunnel. Moreover, its applicability extends to similar environments, showcasing its potential as a valuable resource for hazard assessment and planning in comparable terrains. The deviation road, as well as tunnel T01 of the A1 highway, is in a state of proven instability. It is certain that they will experience continuous, recurrent, and intense landslides. A radical solution to all the geotechnical issues plaguing this section of the highway is to reroute it far away from the Subnumidian formations. A deeper comprehension of the geological and geotechnical intricacies in challenging terrains can significantly enhance the safety and reliability of transportation networks in these demanding contexts.

GIS-based landslide and rockfall susceptibility zoning in Chepelarska River Basin (Western Rhodope Mountains, Bulgaria)

Landslides and rockfalls are one of the most widespread hazardous phenomena in the Western Rhodopes, Bulgaria. These mass-wasting types occur together with other dangerous phenomena such as river bank erosion and soil degradation, and can cause a lot of infrastructural damage. The main purpose of this paper is to perform GIS-based landslide and rockfall susceptibility zoning and to elaborate susceptibility maps at a scale of а river basin on the example of the Chepelarska River Basin (RB). In order to achieve this goal, nine causal factors have been selected, and map of landslides and field study have been used. These factors were reclassified using histograms and the natural breaks method. In addition, the frequency ratio values of each class have been calculated. To obtain the causal factors impact on sliding and rockfalls, the weights and weighted averages sum of their classes were calculated and analysed. For the complex analysis and elaboration of mass-wasting susceptibility maps, weighted overlay and weighted sum methods have been applied. The obtained results show that from the total 38 landslides and rockfalls in the study area, 26 of them fall into the very high susceptibility zone applying a weighted overlay, and 25 of them fall into the same zone using the weighted sum method. These results and the used methods could be used for other small and medium-sized river basins.

Landslide susceptibility mapping for the Red Sea Mountains: A multi-criteria decision analysis approach

The Earth's climate system is clearly warming. Unquestionably, climate change has an impact on landslides by affecting the stability of both naturally occurring and artificially constructed slopes. Less certain are the type, extent, degree, and direction of the changes in stability conditions as well as the location, abundance, activity, and frequency of landslides in response to the expected climate changes. Climate change threatens global ecosystems, causing severe human and economic losses through sea level rise, hurricanes, flooding, landslides, and heat waves. Landslides in high mountains could become more common as a result of these changes. The high-speed electric train project in Egypt seeks to provide sustainable transportation, lessen climate change, and enhance public health. The study area was conducted along a 70-km track in the Eastern Desert, Egypt, across steep and rugged basement rocks of the Red Sea mountain chain (RSM). The geographic pinpointing of landslide susceptibility zones plays a significant role in promoting hazards mitigation in mountainous areas, such as RSM. The aim of this research project is to create a landslide susceptibility map (LSM) utilizing multi-criteria decision analysis (MCDA) in the region using information value (IV) and analytical hierarchy process (AHP) approaches that encompass the third proposed track of the high-speed electric railroad Qena-Hurghada between the RSM. The landslide inventory map, which has a total of 12 single landslide locations, was created based on fieldwork using hand-held GPS and Landsat 8 images. The two decision-making techniques were used to contrast the LSM factors based on the knowledge of experts using the qualitative method of AHP with the statistical approach that depends on the field data IV. Ten landslide-influencing parameters taken into consideration are slope, lithology, rainfall distribution, distance to fault, distance to track, seismicity of area, distance to epicenter, land cover, distance to stream, and aspect. The parameters were obtained using LANDSAT8 and SRTM DEM of a 30 m resolution, vectorized lithological structure units from a 1:2000, 000 map, and data between (1913–2000) for yearly precipitation and (1997–2018) for seismic activity of the research area. A GIS-based method produces a themed layer map for each geoenvironmental element, weighing each parameter after examination. A LSM was produced using a combination of thematic layers that were then separated into four grades: low, moderate, high, and very high, based on the intensity of the landslide vulnerability. According to AUC data, the IV model has a better success rate 0.85 than the AHP model 0.82. The study's models, especially when applied to the IV method, showed good accuracy in forecasting landslide vulnerability. The method shows rising landslide percentages with susceptibility and delivers accurate results that are compatible with the landslide inventory map. Environmental decisions often lead to unintended consequences due to lost information, incorrect factor selection, limited expert expertise, and ignoring uncertainty. This investigation emphasizes the importance of LSM as a tool for decision-makers to minimize landslide losses and damage.

GIS-Based Landslide Susceptibility Zonation Mapping Using the Weighted Information Model in Erlang Mountain - Zheduo Mountain Power Transmission Project, China

GIS-Based Landslide Susceptibility Zonation Mapping Using the Weighted Information Model in Erlang Mountain - Zheduo Mountain Power Transmission Project, China

High resolution landslide susceptibility mapping using ensemble machine learning and geospatial big data

Landslide susceptibility represents the potential of slope failure for given geo-environmental conditions. The existing landslide susceptibility maps suffer from several limitations, such as being based on limited data, heuristic methodologies, low spatial resolution, and small areas of interest. In this study, we overcome all these limitations by developing a probabilistic framework that combines imbalance handling and ensemble machine learning for landslide susceptibility mapping. We employ a combination of One -Sided Selection and Support Vector Machine Synthetic Minority Oversampling Technique (SVMSMOTE) to eliminate class imbalance and develop smaller representative data from big data for model training. A blending ensemble approach using hyperparameter tuned Artificial Neural Networks, Random Forests, and Support Vector Machine, is employed to reduce the uncertainty associated with a single model. The methodology provides the landslide susceptibility probability and a landslide susceptibility class. A thorough evaluation of the framework is performed using receiver operating characteristic curves, confusion matrices, and the derivatives of confusion matrices. This framework is used to develop India's first national-scale machine learning based landslide susceptibility map. The landslide database is carefully curated from global and local inventories, and the landslide conditioning factors are selected from a multitude of geophysical and climatological variables. The Indian Landslide Susceptibility Map (ILSM) is developed at a resolution of 0.001° (∼100 m) and is classified into five classes: very low, low, medium, high, and very high. We report an accuracy of 95.73 %, sensitivity of 97.08 %, and matthews correlation coefficient (MCC) of 0.915 on test data, demonstrating the accuracy, robustness, and generalizability of the framework for landslide identification. The model classified 4.75 % area in India as very highly susceptible to landslides and detected new landslide susceptible zones in the Eastern Ghats, hitherto unreported in the government landslide records. The ILSM is expected to aid policymaking in disaster risk reduction and developing landslide prediction models.

Delineation of landslide susceptible zones using Frequency Ratio (FR) and Shannon Entropy (SE) models in northern Rif, Morocco

Delineation of landslide susceptible zones using Frequency Ratio (FR) and Shannon Entropy (SE) models in northern Rif, Morocco

Identification of Urban Centre and Rural Growth Centres Around Guwahati and Its Surrounding Rural Region Using Hierarchical Settlements, Nested Hexagons, Remote Sensing and GIS

Guwahati city is the highest order urban center of Assam and is an important gateway to the north eastern region of India.  In this study, a 50km buffer from the master plan boundary of Guwahati Metropolitan Development Authority (GMDA) is selected for identifying potential urban centers and rural growth centers (URGC) of different order for decentralized planning and inter and intra-administrative cooperation around the city using multi-parametric criteria. This includes central place theory, nested hexagon method and thematic information on groundwater potential zones, land use/land cover, flood prone and landslide susceptible zones. Out of the 32 identified potential villages, 15 are proposed for new urban centers and 17 are proposed for development as rural growth centers. 9 towns are also proposed for up-gradation to higher order for proper spatio-functional interaction. However, several suggestions and preventive measures were made before initiating developmental expansion which needs to be considered. The findings of this study would be useful for decentralized planning to minimize the economic imbalances, rural migration and sustainable development of the region.

Assessing the Susceptibility of Landslides in the Tuoding Section of the Upper Reaches of the Jinsha River, China, Using a Combination of Information Quantity Modeling and GIS

Combined with visible light remote sensing technology and InSAR technology, this study employed the fundamental principles of the frequency ratio model, information content model, and analytic hierarchy process to assess the susceptibility of the study area. Nine susceptibility assessment factors such as elevation, slope, aspect, water system, vegetation coverage, geological structure, stratum lithology, rainfall, and human activities were selected, and the factor correlation degree was calculated by using the relative area density value of the landslide. The frequency ratio model and information content model were selected to carry out landslide susceptibility zoning, and the accuracy of the two models was verified by the ROC curve and density method. The results indicate that the information content model performed relatively well. Therefore, the information model, combined with the analytic hierarchy process and fuzzy superposition method using the landslide point density map, was chosen to evaluate landslide susceptibility. The study area was divided into five levels of landslide hazard, ranging from low to high, using the natural discontinuity point method. The results show that the area of each hazard zoning is 197.48, 455.72, 408.21, 152.66, and 16.22 km2 from low to high, and the proportion of landslides in the corresponding area is 0.17%, 1.60%, 3.88%, 8.41%, and 16.65%, respectively. It can be seen that with the increase in the hazard level, the proportion of landslides also increases significantly, which verifies the accuracy of the hazard results. Additionally, four representative landslides in the study area were selected for analysis to understand their characteristics and underlying mechanisms. The results revealed that these landslides were notably influenced by the density of the Jinsha River and the surrounding roads. The susceptibility assessment outcomes for geological disasters align well with the current situation of landslide occurrences in the Tuoding river section, demonstrating high accuracy. This study provides a scientific foundation for effective prevention and control measures against local landslide disasters.

Forensic investigation of the Gouache landslide disaster, Western Region, Cameroon

This study made use of a disaster forensic approach to investigate the root causes of the landslide disaster which occurred in Gouache,  West Region, Cameroon on 29th of October, 2019 that registered 49 to 60 deaths. The aims of the study were to;1) identify the natural  and man-made causes of the landslide; 2) determine household vulnerability factors; and 3) identify the administrative weaknesses that  led to the disaster. To achieve this, geological data was collected from field observations; rainfall data for a period of 30 years (1988 to  2018) with additional monthly data (January to December) for 2019 was obtained from the National observatory on Climate change  (ONACC), Yaounde; satellite imagery from Landsat ETM+ and 8 (2001 and 2020) and SRTM DEM were used to assess land use/ landcover  changes and slope respectively; questionnaire administration, household and key informant interviews were used as primary data  collection tools. Findings revealed that the root causes were: heavy rainfall that occurred prior to and during this event; the presence of  weathered granitic rocks on steep slopes, and springs that increased landslide susceptibility. Susceptibility to landslides was fostered by  human activities such as: significant reduction of the savannah vegetation and cutting of the slope for house construction. Household  vulnerability factors were lack of awareness of landslide hazards and large household sizes (4-6 persons). Over 70 % of the landowners  constructed without a building permit and had no land certificate. The administrative authorities acknowledged that Gouache was officially recognized as a landslide susceptible zone. The findings generated in this study would be useful for the government to  strengthen or re-adjust existing policies or strategies of disaster risk reduction in the area.  

Comparative analysis of certainty factor and analytic hierarchy process for landslide susceptibility zonation in parts of Solan, Himachal Pradesh, India

The state of Himachal Pradesh in India is one of the most important hotspots when it comes to landslides; and Kandaghat, a tehsil in the Solan district of Himachal Pradesh having religious and tourism importance, is substantially affected by frequent landslides causing road blocking. In the present study, the analytic hierarchy process (AHP) and certainty factor (CF) techniques, which form part of the geographic information system (GIS)-based landslide susceptibility models, were used to prepare a landslide susceptibility map for the Kandaghat region, for which, as a preliminary step, an inventory of 214 live landslides was prepared from the Bhukosh data directory. The landslide inventory was cross-verified on the Google Earth platform. About nine landslide causative factors (slope, curvature, aspect, soil, rainfall, land use–land cover, lithology, drainage density and lineament density) were considered for the study area, and against the backdrop of these, the corresponding thematic maps were prepared and used in turn for the preparation of the final landslide susceptibility map. Based on the two mentioned techniques, the thematic maps were assigned weights according to their prominence and dynamic processes in the study area. The model performance for each method was evaluated using the area under the curve (AUC), and the accuracies for the AHP and CF were ascertained as, respectively, 81% and 85.6%. The Himalayan terrains are significantly prone to landslides, and this study outlines the characteristics of one of the important Himalayan towns in terms of vulnerability for landslides, together with providing its classification in terms of slope deformation susceptibility; this procedure can help direct attention towards areas needing to be classified under high to very high landslide susceptibility zones.

Investigation of Landslide Susceptibility Decision Mechanisms in Different Ensemble-Based Machine Learning Models with Various Types of Factor Data

Machine learning (ML)-based methods of landslide susceptibility assessment primarily focus on two dimensions: accuracy and complexity. The complexity is not only influenced by specific model frameworks but also by the type and complexity of the modeling data. Therefore, considering the impact of factor data types on the model’s decision-making mechanism holds significant importance in assessing regional landslide characteristics and conducting landslide risk warnings given the achievement of good predictive performance for landslide susceptibility using excellent ML methods. The decision-making mechanism of landslide susceptibility models coupled with different types of factor data in machine learning methods was explained in this study by utilizing the Shapley Additive exPlanations (SHAP) method. Furthermore, a comparative analysis was carried out to examine the differential effects of diverse data types for identical factors on model predictions. The study area selected was Cenxi, Guangxi, where a geographic spatial database was constructed by combining 23 landslide conditioning factors with 214 landslide samples from the region. Initially, the factors were standardized using five conditional probability models, frequency ratio (FR), information value (IV), certainty factor (CF), evidential belief function (EBF), and weights of evidence (WOE), based on the spatial arrangement of landslides. This led to the formation of six types of factor databases using the initial data. Subsequently, two ensemble-based ML methods, random forest (RF) and XGBoost, were utilized to build models for predicting landslide susceptibility. Various evaluation metrics were employed to compare the predictive capabilities of different models and determined the optimal model. Simultaneously, the analysis was conducted using the interpretable SHAP method for intrinsic decision-making mechanisms of different ensemble-based ML models, with a specific focus on explaining and comparing the differential impacts of different types of factor data on prediction results. The results of the study illustrated that the XGBoost-CF model constructed with CF values of factors not only exhibited the best predictive accuracy and stability but also yielded more reasonable results for landslide susceptibility zoning, and was thus identified as the optimal model. The global interpretation results revealed that slope was the most crucial factor influencing landslides, and its interaction with other factors in the study area collectively contributed to landslide occurrences. The differences in the internal decision-making mechanisms of models based on different data types for the same factors primarily manifested in the extent of influence on prediction results and the dependency of factors, providing an explanation for the performance of standardized data in ML models and the reasons behind the higher predictive performance of coupled models based on conditional probability models and ML methods. Through comprehensive analysis of the local interpretation results from different models analyzing the same sample with different sample characteristics, the reasons for model prediction errors can be summarized, thereby providing a reference framework for constructing more accurate and rational landslide susceptibility models and facilitating landslide warning and management.