Landslide Susceptibility Models Research Articles

Hybrid meta-heuristic machine learning methods applied to landslide susceptibility mapping in the Sahel-Algiers

Landslides are considered as one among many phenomena jeopardizing human beings as well as their constructions. To prevent this disastrous problem, researchers have used several approaches for landslide susceptibility modeling, for the purpose of preparing accurate maps marking landslide prone areas. Among the most frequently used approaches for landslide susceptibility mapping is the Artificial Neural Network (ANN) method. However, the effectiveness of ANN methods could be enhanced by using hybrid metaheuristic algorithms, which are scarcely applied in landslide mapping. In the current study, nine hybrid metaheuristic algorithms, genetic algorithm (GA)-ANN, evolutionary strategy (ES)-ANN, ant colony optimization (ACO)-ANN, particle swarm optimization (PSO)-ANN, biogeography based optimization (BBO)-ANN, gravitational search algorithm (GHA)-ANN, particle swarm optimization and gravitational search algorithm (PSOGSA)-ANN, grey wolves optimization (GWO)-ANN, and probability based incremental learning (PBIL)-ANN have been used to spatially predict landslide susceptibility in Algiers’ Sahel, Algeria. The modeling phase was done using a database of 78 landslides collected utilizing Google Earth images, field surveys, and six conditioning factors (lithology, elevation, slope, land cover, distance to stream, and distance to road). Initially, a gamma test was used to decrease the input variable numbers. Furthermore, the optimal inputs have been modeled by the mean of hybrid metaheuristic ANN techniques and their performance was assessed through seven statistical indicators. The comparative study proves the effectiveness of the co-evolutionary PSOGSA-ANN model, which yielded higher performance in predicting landslide susceptibility compared to the other models. Sensitivity analysis using the step-by-step technique was done afterward, which revealed that the distance to the stream is the most influential factor on landslide susceptibility, followed by the slope factor which ranked second. Lithology and the distance to road have demonstrated a moderate effect on landslide susceptibility. Based on these findings, an accurate map has been designed to help land-use managers and decision-makers to mitigate landslide hazards.

Evaluation of Different Landslide Susceptibility Models for a Local Scale in the Chitral District, Northern Pakistan.

This work evaluates the performance of three machine learning (ML) techniques, namely logistic regression (LGR), linear regression (LR), and support vector machines (SVM), and two multi-criteria decision-making (MCDM) techniques, namely analytical hierarchy process (AHP) and the technique for order of preference by similarity to ideal solution (TOPSIS), for mapping landslide susceptibility in the Chitral district, northern Pakistan. Moreover, we create landslide inventory maps from LANDSAT-8 satellite images through the change vector analysis (CVA) change detection method. The change detection yields more than 500 landslide spots. After some manual post-processing correction, the landslide inventory spots are randomly split into two sets with a 70/30 ratio for training and validating the performance of the ML techniques. Sixteen topographical, hydrological, and geological landslide-related factors of the study area are prepared as GIS layers. They are used to produce landslide susceptibility maps (LSMs) with weighted overlay techniques using different weights of landslide-related factors. The accuracy assessment shows that the ML techniques outperform the MCDM methods, while SVM yields the highest accuracy of 88% for the resulting LSM.

Predicting Landslides Susceptible Zones in the Lesser Himalayas by Ensemble of Per Pixel and Object-Based Models

Landslide susceptibility is a contemporary method for delineation of landslide hazard zones and holistically mitigating the future landslides risks for planning and decision-making. The significance of this study is that it would be the first instance when the ‘geon’ model will be attempted to delineate landslide susceptibility map (LSM) for the complex lesser Himalayan topography as a contemporary LSM technique. This study adopted the per-pixel-based ensemble approaches through modified frequency ratio (MFR) and fuzzy analytical hierarchy process (FAHP) and compared it with the ‘geons’ (object-based) aggregation method to produce an LSM for the lesser Himalayan Kalsi-Chakrata road corridor. For the landslide susceptibility models, 14 landslide conditioning factors were carefully chosen; namely, slope, slope aspect, elevation, lithology, rainfall, seismicity, normalized differential vegetation index, stream power index, land use/land cover, soil, topographical wetness index, and proximity to drainage, road, and fault. The inventory data for the past landslides were derived from preceding satellite images, intensive field surveys, and validation surveys. These inventory data were divided into training and test datasets following the commonly accepted 70:30 ratio. The GIS-based statistical techniques were adopted to establish the correlation between landslide training sites and conditioning factors. To determine the accuracy of the model output, the LSMs accuracy was validated through statistical methods of receiver operating characteristics (ROC) and relative landslide density index (R-index). The accuracy results indicate that the object-based geon methods produced higher accuracy (geon FAHP: 0.934; geon MFR: 0.910) over the per-pixel approaches (FAHP: 0.887; MFR: 0.841). The results noticeably showed that the geon method constructs significant regional units for future mitigation strategies and development. The present study may significantly benefit the decision-makers and regional planners in selecting the appropriate risk mitigation procedures at a local scale to counter the potential damages and losses from landslides in the area.

Using Fuzzy Neural Networks to Model Landslide Susceptibility at the Shihmen Reservoir Catchment in Taiwan

Machine learning algorithms are commonly employed in landslide susceptibility assessments. Recently, algorithms that utilize artificial intelligence have come into prominence. This study attempts to adapt the most fundamental framework of deep learning and introduces fuzzy theory concepts to analyze landslide susceptibility while updating the network parameters with trial-and-error methods. The final analysis results will compare with those of logistic regression (LR). In order to assess the ability of the model to identify landslides in a more objective way, two typhoon events were used as a training event and a validation event, respectively. The results of the analysis show that the area under the curve (AUC) of the fuzzy neural network (FNN) for the training event is 0.915, but the AUC for the validation event drops to 0.746. Although the results of the FNN for training events were better than those of LR, they did not differ much from those of LR in predicting future events. The reason for this is that the difference between the landslide distributions of the training and validation events is too large, making the model biased in its identification. Overall, FNN is still a recommended method for analyzing landslide potential and can be used as a reference for LR.

Assessment of rainfall-induced landslide susceptibility in Artvin, Turkey using machine learning techniques

In this study, the performances of machine learning models, such as artificial neural networks (ANN), gradient-boosting machines (GBM), random forest (RF) and support vector machines (SVM) in rainfall-induced landslide susceptibility mapping were evaluated. For this purpose, the Arhavi, Hopa and Kemalpaşa districts of Artvin, which is one of the highest rainfall areas in Turkey, were identified as the study area. A landslide inventory comprising 533 landslide polygons (3959 pixels at 10-m resolution) was used; 70% of the pixels showing the landslides were used for training the models and the remaining 30% were used to validate the models. For landslide susceptibility modelling, 13 factors associated with landslides were considered. The area under the receiver operating characteristic curve was found to reveal the predictive capabilities of the models. As a result, the prediction rates of the ANN, SVM, RF and GBM models were found to be 93.8%, 94.8%, 96.1%, and 97%, respectively. According to the results, the GBM outperformed other models.

Landslide Susceptibility Assessment of Heraklion Prefecture in Crete, Greece

The purpose of this particular study is the assessment of landslide susceptibility of Heraklion prefecture area of Crete Island through geospatial analysis in order to present the general susceptibility zonation, so that it may be used as a basis in future extensive and detailed landslide susceptibility research, which will lead to the prevention of possible future landslides. In order to accomplish this task, it is essential to study the geological, geomorphological, geotectonic, hydrological and hydrogeological characteristics of the area, as well as the land use and compare this information with real landslide occurrence data from the study area. Based on this study it became clear that the area is defined by medium to very steep slopes and a geotechnical and geological setting that are extremely favorable for landslides. These reasons, as well as the fact that this prefecture is known to be the most densely populated prefecture of Crete Island created great interest in this area for research of this kind to prevent the loss of human life and the structural damage of properties. In order to study the landslide susceptibility of Heraklion prefecture, firstly, a geodatabase, which includes the landslide inventory, was created. The landslide inventory is consisted of previous landslides, which were collected from related geotechnical reports with all the essential spatial and descriptive information (such as locations and causes of the landslides). Furthermore, the necessary raster and vector data for the geospatial analysis of the area were obtained. Because the landslide susceptibility is a multifactorial problem it was decided to use a relating method to the problem’s nature, which was the methodology of the Analytical Hierarchy Process (AHP), as it is a simple and widely used method for similar studies, which makes it a great tool for this introductory study. In addition, this method was implemented in a GIS environment. For the purpose of the study, ten (10) triggering factors that cause landslides were defined, followed by the creation of a landslide susceptibility assessment map of the study area, depending on the influence of each factor. In accordance with this map, the highest susceptibility mostly corresponds to the mountainous regions of the prefecture. Finally, the success evaluation of the landslide susceptibility model was determined based on the frequency of landslide occurrences of the medium, high, and very high susceptibility zones. Consequently, the model was characterized as 90.48% successful, according to bibliography and the landslide inventory.

Comparative review of data-driven landslide susceptibility models: case study in the Eastern Andes mountain range of Colombia

Comparative review of data-driven landslide susceptibility models: case study in the Eastern Andes mountain range of Colombia

Landslide Susceptibility Mapping along a Rapidly Uplifting River Valley of the Upper Jinsha River, Southeastern Tibetan Plateau, China

As a result of the influence of plate movement, the upper reaches of Jinsha River have strong geological tectonic activities, large topographic fluctuations, and complex climate characteristics, which result in the frequent occurrence of landslide disasters. Hence, there is the need to carry out landslide susceptibility mapping in the upper reaches of Jinsha River to ensure the safety of local people’s property and the safe exploitation of hydraulic resources. In this study, InSAR technology and a field geological survey were used to map the landslides. Then, the curvature watershed method was used to divide the slope units. A conditioning factor system was established, which can reflect the characteristics of the rapid uplift and vertical distribution of rainfall in the special geological environment of the study area. Finally, logistic regression, random forest, and artificial neural network models were used to establish the landslide susceptibility model. The results show that the random forest model is optimal for the landslide susceptibility mapping in this area. Additionally, the area percentages of the very low, low, moderate, high, and very high susceptibility classes were 40.13%, 20.06%, 13.39%, 12.55%, and 13.87%, respectively. Based on the analysis of the landslide susceptibility map, we suggest that the landslide geological hazards resulting from the rapid uplift of the Tibetan Plateau and the significant decrease in sea level during a glacial period in the upper reaches of Jinsha River are controlled by the double disaster effect of the geodynamic system. Consequently, this study can guide local prevention and mitigation.

Landslide susceptibility modeling based on GIS and ensemble techniques

Landslide susceptibility modeling based on GIS and ensemble techniques

Mapping of landslide potential in Pyeongchang-gun, South Korea, using machine learning meta-based optimization algorithms

Landslides are geological hazards that can have severe impacts, threatening both the people and the local environment of highlands or mountain slopes. Landslide susceptibility mapping is an essential tool for predicting landslides and mitigating landslide-associated damage in areas prone to these events. This study aims to investigate the combination of using an adaptive network-based fuzzy inference system (ANFIS) with metaheuristic optimization algorithms: gray wolf optimizer (GWO), particle swarm optimization algorithm (PSO), and the imperialist competitive algorithm (ICA) in mapping landslide potential. The study area was Pyeongchang-gun, South Korea, for which an accurate landslide inventory dataset is available. A landslide inventory map was organized, and the data were separated randomly into training data (70%) and validation data (30%). In addition, 16 landslide-related factors consisting of geo-environmental and topo-hydrological factors were considered as predictive variables. This landslide susceptibility model was be evaluated based on the value of the area under the receiver operating characteristic (ROC) curve (AUC) to measure its accuracy. Based on the maps, the validation results showed that the optimized models of ANFIS-ICA, ANFIS-PSO, and ANFIS-GWO had AUC accuracies of 0.927, 0.947, and 0.968, respectively. The result from the hybrid algorithms model of ANFIS with metaheuristic algorithms outperformed the standalone ANFIS model in terms of accuracy in predicting landslide potential. Therefore, the ML algorithm and optimization algorithm models proposed in this study are more suitable for landslide susceptibility mapping in the study area.

Multi-Temporal Landslide Inventory-Based Statistical Susceptibility Modeling Associated With the 2017 Mw 6.5 Jiuzhaigou Earthquake, Sichuan, China

The 2017 Mw 6.5 Jiuzhaigou earthquake (Sichuan, China) is the first strong ground motion that struck the famous world heritage site, causing widespread landslides and severe rock mass damage effects and landscapes undergoing rapid evolution in the Jiuzhaigou National Geopark. However, the understanding of the variability of pre- and post-earthquake landslide susceptibility and landslide conditioning factor effects over time remains limited. This study aims to carry out multi-temporal statistical landslide susceptibility modeling at the slope-unit level related to this event. To achieve this, we initially used a set of remote sensing imageries in GIS to obtain systematic landslide inventories across the pre-, co-, and post-seismic periods. Based on three landslide inventory datasets, we developed three statistical models by incorporating 14 landslide conditioning (seismic, topographic, and geologic) factors into a binary logistic regression (BLR) model. Finally, we utilized the area under the receiver operating characteristic (AUC) (QA) curve to assess each model’s calibration and validation performance. The results show that the BLR model has good prediction applicability for both normal and seismic landslides in the study area with outstanding to excellent predictive accuracy for Mod1 (pre-seismic, AUC = 0.801), Mod2 (co-seismic, AUC = 0.942), and Mod3 (post-seismic, AUC = 0.880) periods. There are variations in both the importance of landslide conditioning factors and susceptibility maps through time, and the number of slope units with a mean probability over 0.8 from only one (pre-seismic) increased to 21 (post-seismic). The dynamic susceptibility maps are of great significance for identifying potentially unstable slopes and providing references for hazard and risk assessment, which could provide new insights into geo-environmental protection and regional landslide evaluation in scenery spots, even for those world heritage sites in the tectonic active mountainous region. Moreover, more frequent or extended observation periods could contribute a further understanding of the post-seismic landslide developments in the Jiuzhaigou area.

Distinct Susceptibility Patterns of Active and Relict Landslides Reveal Distinct Triggers: A Case in Northwestern Turkey

To understand the factors that make certain areas especially prone to landslides, statistical approaches are typically used. The interpretation of statistical results in areas characterised by complex geological and geomorphological patterns can be challenging, and this makes the understanding of the causes of landslides more difficult. In some cases, landslide inventories report information on the state of activity of landslides, adding a temporal dimension that can be beneficial in the analysis. Here, we used an inventory covering a portion of Northwestern Turkey to demonstrate that active and relict landslides (that is, landslides that occurred in the past and are now stabilised) could be related to different triggers. To do so, we built two landslide susceptibility models and observed that the spatial patterns of susceptibility were completely distinct. We found that these patterns were correlated with specific controlling factors, suggesting that active landslides are regulated by current rainfalls while relict landslides may represent a signature of past earthquakes on the landscape. The importance of this result resides in that we obtained it with a purely data-driven approach, and this was possible because the active/relict landslide classification in the inventory was accurate.

Performance Assessment of Event-Based Ensemble Landslide Susceptibility Models in Shihmen Watershed, Taiwan

While multi-year and event-based landslide inventories are both commonly used in landslide susceptibility analysis, most areas lack multi-year landslide inventories, and the analysis results obtained from the use of event-based landslide inventories are very sensitive to the choice of event. Based on 24 event-based landslide inventories for the Shihmen watershed from 1996 to 2015, this study established five event-based single landslide susceptibility models employing logistic regression, random forest, support vector machine, kernel logistic regression, and gradient boosting decision tree methods. The ensemble methods, involving calculating the mean of the susceptibility indexes (PM), median of the susceptibility indexes (PME), weighted mean of the susceptibility indexes (PMW), and committee average of binary susceptibility values (CA) of the five single models were then used to establish four event-based ensemble landslide susceptibility models. After establishing nine landslide susceptibility models, using each inventory from the 24 event-based landslide inventories or a multi-year landslide inventory, we identified the differences in landslide susceptibility maps attributable to the different landslide inventories and modeling methods, and used the area under the receiver operating characteristic curve to assess the accuracy of the models. The results indicated that an ensemble model based on a multi-year inventory can obtain excellent predictive accuracy. The predictive accuracy of multi-year landslide susceptibility models is found to be superior to that of event-based models. In addition, the higher predictive accuracy of ensemble landslide susceptibility models than that of single models implied that these ensemble methods were robust for enhancing the model’s predictive performance in the study area. When employing event-based landslide inventories in modeling, PM ensemble models offer the best predictive ability, according to the Kruskal–Wallis test results. Areas with a high mean susceptibility index and low standard deviation, identified using the 24 PM ensemble models based on different event-based landslide inventories, constitute places where landslide mitigation measures should be prioritized.

The influences of the spatial extent selection for non-landslide samples on statistical-based landslide susceptibility modelling: a case study of Anhui Province in China

The influences of the spatial extent selection for non-landslide samples on statistical-based landslide susceptibility modelling: a case study of Anhui Province in China

A new approach based on integration of random subspace and C4.5 decision tree learning method for spatial prediction of shallow landslides

The research approaches a new machine learning ensemble which is a hybridization of Random subspace (RS) and C4.5, named RandSub-DT, for improving the performance of the landslide susceptibility model. This is based on the GIS database, including 170 landslide polygons and ten predisposing landslide factors, i.e., slope, aspect, curvature, TWI, land use, distance to road, distance to the river, soil type, distance to fault, and lithology. We carried out this study in the Halong and Cam Pha City areas which are important economic centers in the Quang Ninh province, Vietnam, where landslides seriously influence the daily life of the citizen causing economic damage. We then used a GIS database to construct and validate the proposed RandSub-DT model. The model performance was assessed using a confusion matrix and a set of statistical measures. The result showed that the RandSub-DT model with the classification accuracy of 90.34% in the training dataset and the prediction capability of 77.48% had a high performance for landslide prediction. This research proved that an ensemble of the C4.5 and RS provided a highly accurate estimate of landslide susceptibility in the research area.

Influence of sampling design on landslide susceptibility modeling in lithologically heterogeneous areas

This work aims at evaluating the sensitivity of landslide susceptibility mapping (LSM) to sampling design in lithologically-heterogeneous areas. We hypothesize that random sampling of the landslide absence data in such areas can be biased by statistical aggregation of the explanatory variables, which impact the model outputs. To test this hypothesis, we train a Random Forest (RF) model in two different domains, as follows: (1) in lithologically heterogeneous areas, and (2) in lithologically homogeneous domains of the respective areas. Two heterogeneous areas are selected in Japan (125 km2) and Romania (497 km2), based on existing landslide inventories that include 371 and 577 scarps, respectively. These areas are divided into two, respectively three domains, defined by lithological units that reflect relatively homogeneous topographies. Fourteen terrain attributes are derived from a 30 m SRTM digital elevation model and employed as explanatory variables. Results show that LSM is sensitive to a random sampling of the absence data in lithologically heterogeneous areas. Accuracy measures improve significantly when sampling and LSM are conducted in lithologically homogeneous domains, as compared to heterogeneous areas, reaching an increase of 9% in AUC and 17% in the Kappa index.

Evaluation of Parameter Selection in the Bivariate Statistical-based Landslide Susceptibility Modeling (Case Study: the Citarik Sub-watershed, Indonesia)

A landslide susceptibility mapping is essential for landslide hazard mitigation to reduce the associated risk. This paper aims to present the results of the landslide susceptibility modeling in the Citarik sub-watershed using three bivariate statistical-based methods, i.e., frequency ratio (FR), information value (IV), and weight of evidence (WoE). The main objective of this study is to evaluate the significance of the threshold of the area under curve (AUC) value in parameter selection. In this study, 118 landslide pixels were compiled from Google Earth images, unmanned aircraft vehicle (UAV) aerial photos taken just after the landslide, official landslide reports, and field observation. Thirteen landslide causative factors were prepared in Geographic Information System (GIS) environment, derived from various satellite images and maps. The landslide data were divided into two groups, 70% of data as training data and the rest as test data. Two scenarios that involve a different number of parameters were compared to explain the threshold of the AUC value in parameter selection and model accuracy. The result of this study shows that the AUC value threshold of 0.6 for parameter selection cannot be applied in all cases, and the performance of both two scenarios was excellent in assessing landslide susceptibility in this study area. Those three landslide susceptibility zonation maps of the best scenario showed that the sub-watershed's northern, northeastern, south-eastern, and southern parts were under high to very high susceptibility to landslides, including the Cimanggung area where a recent deadly double landslide occurred.

Modelling and mapping of landslide susceptibility regulating potential ecosystem service loss: an experimental research in Saudi Arabia

The study aims to create a novel artificial intelligence model-based landslide susceptibility model (LSM) at Aqabat, Saudi Arabia. For LSM, a combination of bagging, dagging, random forest (RF) ensemble with locally weighted learning (LWL), viz. bagging-LWL, dagging-LWL, and RF-LWL has been developed. The 50 landslide areas were divided into two categories training (40) and testing (10). For training datasets, the LWL-Bagging model had the highest AUC value of ROC curve (AUC-0.91), followed by LWL-RF (AUC-0.881), LWL-Dagging (AUC-0.88) and LWL (AUC-0.875). For testing datasets, AUC values of ROC curve of 0.891, 0.876, 0.868, and 0.844 were found for LWL-Bagging, LWL-RF, LWL-Dagging, and LWL, respectively. Based on AUC value, all hybrid machines learning models performed well, but LWL-Bagging model performed as the best fit model. The potential ecosystem loss services were assessed by assessing possible landslide hazards. The research delivers reliable results to influence policy decisions on ecosystem management and landslide-assessment.

Landslide susceptibility modeling based on remote sensing data and data mining techniques

Landslide susceptibility modeling based on remote sensing data and data mining techniques

Improved tree-based machine learning algorithms combining with bagging strategy for landslide susceptibility modeling

Improved tree-based machine learning algorithms combining with bagging strategy for landslide susceptibility modeling