Search

This paper evaluates a comparison between three machine learning algorithms (MLAs), namely support vector machine (SVM), multilayer perceptron artificial neural network (MLP-ANN) and random forest (RF), in landslide susceptibility mapping and addresses a optimization algorithm to optimize the performance of a MLA to yield more accurate and reliable results. A genetic algorithm (GA) approach as a part of evolutionary algorithms was utilized in order to optimize the performance of best model among three utilized MLAs. The study area (Tarom–Khalkhal sub-basin) is located in North-West of Iran with mountainous nature (western part of Alborz Mountains), wherein numerous landslide occurrences were recorded. In this case, fifteen predisposing factors, gathered from aerial images and field surveys, were considered to generate the final landslide susceptibility models. The validation procedure was conducted with taking advantage of confusion matrices for different algorithms. Finally, landslide susceptibility maps were generated and evaluated through receiver operator characteristic (ROC) curves. RF algorithm showed the best performance; therefore, hybridized genetic random forest (GRF) was employed in order to optimize the hyperparameters (number of trees, number splits and depth) of the model, which can affect the performance of model. As a result, GRF has best performance among all mentioned algorithms with AUC = 0.93. As a conclusion, genetic algorithm was found to be suitable in optimizing the performance of machine learning algorithms, which is crucial when it comes to landslide susceptibility mapping. [ABSTRACT FROM AUTHOR]

Landslide susceptibility analysis based on the strong ability of data mining of Geographic Information System (GIS) has become a hot topic in international landslide research. This paper used optimized decision tree and GIS databases to analyze the sensitivity in the northwest mountain areas of Yunnan province of China, and then discussed the formation mechanism of the landslide happened in the area. The translational landslide located in the area with an average gradient less than or equal to 28.7° was reclassified as a higher level 3 sensitive area than before according to the normalized different fault index (NDFI). The results showed that the data mining based on GIS 3D space–time information database can help to find the unique topography, geology hydrology and the other typical spatial information of some special typed of landslides such as translational landslides, thus it can illustrate the relationship between the landslides and their sensitivity factors. The improved landslide susceptibility analysis will provide a new method for identifying the genetic mechanism of landslide, and play an important role in the government regional planning and disaster prevention measures. [ABSTRACT FROM AUTHOR]

Estimating the likelihood of landslides has proven to be critical for development and protection of infrastructure (e.g. pipelines, roads) and urban settlements. Currently, for regional studies of landslide susceptibility only qualitative or statistical evaluations are possible due to the large spatial variability of geological properties, topography, rainfall patterns, etc. In this paper, we explore an alternative to these approaches using data-driven methodologies to determine landslide susceptibility. We give special attention to the use of geographical information systems, machine learning and statistical techniques to build landslide susceptibility maps. These methods have input as fourteen key causative factors that might influence landslides occurrence. Additionally, feature extraction and feature selection are performed to evaluate if dimensionality reduction increases the prediction accuracy of the machine learning models. The models were compared using a case study in the Eastern Cordillera of Colombia, where the best performing model achieved a predictive performance of 93.07 % . [ABSTRACT FROM AUTHOR]

Long-runout landslides can be extremely hazardous and unpredictable. The prediction of long-runout landslides is challenging because it is unclear what factors control their mobility, and it is unclear how to best measure their runout given that they traverse changing topography. In this paper, we document the development and statistical evaluation of a Landslide Runout Score (LRS) system to predict short, medium, and long runout using a new mobility measure, the unitless Runout Number L/A1/2 (where L is landslide length and A is landslide area). The Runout Number has previously been correlated to three geomorphological factors (planimetric curvature, sand content, and upslope contributing area normalized to landslide area), so the LRS uses these factors as inputs. These factors are readily calculated using geographic information systems (GIS) and publicly available data sources. The LRS system predicts runout categories (short, medium, and long) with one-vs.-all accuracies (i.e., percentage of correct predictions within a given class) of 75, 58, and 72%, respectively, for a total weighted accuracy of approximately 65%. The results of this work are summarized in a worksheet that can be used by geologists and engineers to develop preliminary predictions of landslide runout behavior, which can be incorporated into or used alongside regional-scale landslide hazard assessments. [ABSTRACT FROM AUTHOR]

Landslides are a serious natural hazard in the world. A map of landslide susceptibility can help to effectively reduce losses. In this paper, a hybrid ensemble technique based on random subspace (RS) and a J48 decision tree named RS–J48T was proposed for landslide susceptibility mapping. This model could enhance the effect of a single classifier significantly and solve the problem of overfitting. Qingchuan County, Sichuan (China) was taken as a study area. A geospatial database which consisted of 640 landslide locations and 12 factors was constructed for this study. The J48 decision tree, artificial neural network (ANN), and other ensemble techniques like AdaBoost and Bagging, were selected for comparison. Receiver operating curves and some statistical indices were used for model validation. The results showed that the RS–J48T model had the better fitting capability (AUC = 0.875), and the best prediction capability (AUC = 0.769) compared to other models. Overall, the novel hybrid model could be a promising way for generating landslide susceptibility maps for other prone areas. [ABSTRACT FROM AUTHOR]

Geomorphological analysis of landslide processes in mountainous terrains with difficult access has benefited from virtual representation of topography through the use of high-resolution digital elevation models (DEMs) generated by use of light detecting and ranging. Generic models of overlay and interpolation take advantage of the high-resolution DEMs for volume calculation. However, the advantages and limitations of the models at regional scale have received scant attention. These challenges are magnified in landslide hazard zonation mapping projects at state or national level, where the models need to be implemented for large datasets. To address the above deficiency, this paper presents a means to estimate landslide volume production and distribution by taking full advantage of LiDAR and to standardize landslide volume calculation in a geographic information system (GIS). We implemented two generic landslide volume models by using Python scripts; this is a systematic methodology for modeling volume of shallow and deep-seated landslides. The models were tested in real and theoretical conditions to highlight advantages and limitations. At the same time, we explored how the interpolation model is affected by local altimetric variation. The results show that one of the models can be used to make first-order interpretations regarding volume of eroded debris for landslide deposits at a local or national scale, while the other can help to assess the sequence of landslide activity. Theoretical evaluations show that local altimetric variation of < 1 m could lead to errors of almost 17%. The approach is explored with examples from Sumas Mountain in Whatcom County, Washington, USA. [ABSTRACT FROM AUTHOR]

Landslides are among the most destructive natural hazards causing loss of life, destruction of infrastrucures and damage to properties, especially in hilly and mountaneous areas all over the world. To properly plan and manage these regions, it is crucial to accurately identify areas susceptible to landslides. In this study, we developed and applied novel hybrid machine learning models, RF-FPA and MAB-FPA, which combine forest by penalizing attribute (FPA), with two ensemble techniques, random forest (RF) and multiboost (MAB), to construct landslide susceptibility maps of Dien Bien province in Vietnam. For the development of hybrid models, we have used data of 665 landslide events and 12 landslide conditioning factors namely distance to rivers, elevations, distance to roads, normalized difference vegetation index, faults, curvature, slope, flow accumulation, stream power index, geology, topographic wetness index (TWI), and aspect. Standard statistical measures including area under the receiver operating characteristic (AUC-ROC) curve were used to evaluate predictive performance of the developed models. Results showed that both proposed novel hybrid models performed well in the correct identification of landslide susceptible areas, with RF-FPA (AUC = 0.840) slightly outperforming MAB-FPA (AUC = 0.814). The performance of the proposed hybrid models was also significantly better than that of single base classifier FPA (AUC = 0.807). In conclusion, while all the studied models performed well, the novel RF-FPA model is a more promissing tool for accurate mapping and correct prediction of landslide susceptibile areas compared to MAB-FPA amd FPA models. This approach of integrating RF with FPA can also be applied to spatially predict landslides in other areas, considering local geo-environmental conditions for proper management of landslide-prone areas. [ABSTRACT FROM AUTHOR]

The present work aimed to assess the feasibility of three methods, namely, analytical hierarchy process, frequency ratio and binary logistic regression, that are used for landslide susceptibility analysis. Twelve causal factors such as slope, relative relief, road buffer, fault/fold and thrust (FFT) buffer, topographic wetness index (TWI), aspect, stream power index (SPI), drainage buffer, profile curvature, land used/land cover (LULC), soil and geology as well as a landslide inventory were retrieved using multi-source spatial data. Correlation between landslides and causative factors was estimated using frequency ratio on a geospatial platform with 234 landslide locations, extracted using field visits and remote sensing imageries and found to be reasonably influencing landslides in the study region. These factors were then used in all three landslide susceptibility models to calculate and generate a landslide susceptibility map (LSM). LSM of the study region Mao-Maram (tehsil) of Manipur was generated with five classes: very low, low, medium, high, and very high. Cumulative percentage curve and receiver operating characteristic (ROC) approaches were used for validation of the model performance. Cumulative percentage curve area under curve (AUC) accuracy for frequency ratio, analytical hierarchy process and binary logistic regression were found to be 77.24%, 73.22%, and 80.32%, respectively. ROC curve under AUC for frequency ratio 88.7%, analytical hierarchy process 77.4%, and binary logistic regression 90%. A landslide susceptibility map derived from every three models can be helpful for city or village planners for planning, development, mitigation, and disaster prevention. [ABSTRACT FROM AUTHOR]

A landslide susceptibility analysis has been made in the Río Aguas catchment (Almeria, Southeast Spain), using two statistical models, Maximum Entropy (MaxEnt) and Geographically Weighted Logistic Regression (GWLR). For this purpose, a previous landslide inventory has been used and re-elaborated, reaching a total incidence of 2.58% of the whole area. Different types of movements have been distinguished, being rock falls, slides and complex movements the predominant. From the inventory, the centroid of the rupture zone has been extracted to represent the landslides introduced in the models. A previous factor analysis has been made, using 12 predictors related to morphometry, hydrography, geology and land cover, with 5 m grid spacing, allowing the selection of factors to be used in the analysis and discarding those showing correlation between them. Then, MaxEnt and GWLR models are applied using different distributions of training and testing samples from the landslide inventory. For the validation, the Area Under the Curve of the Receiver Operating Characteristic (AUC-ROC) has been used but additionally, the degree of fit (DF) has allowed to validate the rupture zones themselves, not only the centroids. Results show an excellent prediction with both metrics in all the methods and samples, but the better results are obtained in the GWLR method for AUC and in the MaxEnt for the degree of fit. Therefore, a consensus model of both methods has been obtained, that improves even more the results reaching an AUC value of 0.99 and a degree of fit of 90%. [ABSTRACT FROM AUTHOR]

This study aimed to examine the influence of the random selection of landslide training and testing sets on the predictive performance of the shallow landslide susceptibility modelling at regional scale. The performance of frequency ratio (FR), information value (IV), logistic regression (LR), and maximum entropy (ME) methods were tested and compared for modeling shallow landslide susceptibility in Calabria region, (South Italy). A landslide database of 22,028 shallow landslides, randomly split into training (70%) and testing (30%) sets, was combined with 15 predisposing factors (lithology, soil texture, soil bulk density, soil erodibility, drainage density, land use, elevation, local relief, slope gradient, slope aspect, plan curvature, topographic wetness index, stream power index, topographic ruggedness index, and topographic position index) to calibrate and validate the models. The robustness of the models in response to changes in the landslide dataset was explored through ten training and test sets replicates. The performance of these models was evaluated using several statistical indices and the ROC curve method. The results showed that all the four methods applied achieve promising performance on the prediction of shallow landslide susceptibility at regional scale. The comparison between four methods displayed that the ME is the best performing (AUC = 0.866), followed by the LR (AUC = 0.845), FR (AUC = 0.813), and finally IV (AUC = 0.800). In addition, the findings showed that the accuracy of the four methods for modeling shallow landslide susceptibility was quite robust when the training and testing sets were changed (i.e. a very low sensitivity to varying training/testing sets). [ABSTRACT FROM AUTHOR]

This effort is made to develop a landslide hazard map and perform the kinematic slope instability analysis of the critical sections at the Dasu dam site and its reservoir area. The study area is located at the Indus River in the Kohistan District within the highly elevated Himalayan terrain. The fragile geologic setup and high tectonic activity make this region favorable to produce frequent slope failures and landsliding events. First, a regional level landslide inventory map was created using GIS and remote-sensing tools. For this a hill shaded topographic map was created by overlapping a topographic map on to a hillshade image extracted from satellite based PALSAR Digital Elevation Model (DEM) with 12.5 m spatial resolution in GIS environment. Numerous past landsides were delineated using the expert driven knowledge of anomalous topographic expression like, crenulated contours, divergent contours, isolated topographic knobs, extended topographic ridges and arcuate headscarf evacuation areas. Total 120 landslides were marked, later a field visit was carried out for the validation of this exercise. The overall efficiency of this work was 85% which indicates that the GIS and remote sensing tools utilized for this study are reliable and effective in producing regional landslide inventory maps. Moreover, the scanline surveys were conducted at few selected locations to generate discontinuities data for the evaluation of slope instability issues by using kinematic analysis in DIPS 7.0 software. The study successfully highlights the usefulness of GIS and RS tools and in combination with kinematic analysis for landslides hazard assessment for mega projects. The outcomes of this research could serve as a guide in risk assessment and mitigation work in future to stabilize the critical slopes near Dasu dam site and along its reservoir area. [ABSTRACT FROM AUTHOR]

Landslides are one of the most destructive natural phenomena in the world, which occur mostly in mountainous areas and cause damage to the economic sectors, agricultural lands, residential areas and infrastructures of any country, and also threaten the lives and property of human beings. Therefore, landslide susceptibility mapping (LSM) can play a critical role in identifying prone areas and reducing the damage caused by landslides in each area. In the present study, deep learning algorithms including convolutional neural network (CNN) and long short-term memory (LSTM) were used to identify landslide prone areas in Ardabil province, Iran. Then 312 landslide locations were identified and randomly divided into train and test datasets, and according to previous studies and environmental conditions in the study area, twelve factors affecting the occurrence of landslides were selected. The ratio of the importance of each influential factor in landslide occurrence was obtained through information gain ranking filter method and it was found that land-use and profile curvature had the highest and lowest impacts, respectively. Afterwards, LSMs were generated using CNN and LSTM algorithms. In the next step, the performance of the models was evaluated based on the area under curve (AUC) value of receiver operating characteristics curve and the root mean square error (RMSE) method. The AUC values for CNN and LSTM models were 0.821 and 0.832, respectively. Furthermore, the RMSE values in the CNN model for each of the training and testing dataset were 0.121 and 0.132, respectively. The RMSE values in the LSTM model for each of the training and testing dataset were 0.185 and 0.188, respectively. Therefore, it can be concluded that LSTM performance is slightly better than CNN; but in general, both models have close performance and the accuracy of both models is acceptable. [ABSTRACT FROM AUTHOR]

In Nefas-Mewcha to Weldiya road corridor (study area), landslide incidence resulted in the death of people, devastation of infrastructure, properties, crops, and agricultural lands. To reduce damages due to landslide incidences, a complete landslide susceptibility mapping was carried out using GIS-based index of entropy (IOE) and frequency ratio (FR) models. Detailed fieldwork and google earth imagery analysis were used to identify 712 landslides. These landslides were divided into two categories: 498 (70%) for modeling and 214 (30%) for model validation. The spatial relationship between pre-existing landslides and 12 landslide factors was performed. Using a raster calculator, the weighted landslide factors were combined to provide a landslide susceptibility index (LSI). The natural break classification method was used to divide the LSI into five categories: very low, low, moderate, high, and very high susceptibility zones. The area under the curve (AUC) and the receiver operating characteristic (ROC) curves were used to assess the models' performance and accuracy. The results showed that the IOE model (AUC = 70%) performed somewhat better than the FR model (AUC = 66.41%) in terms of prediction. The IOE method also showed slightly high model performance compared to FR with the success rate of AUC values (71.3% for IOE and 69% for FR). In the IOE model which was produced after selecting the landslide factors, the success rate showed an increment from 71.3% to 74.5%. Similarly, the FR model also showed significant change in a success rate of 78.1% and a predictive rate of 73.5%. According to this finding, the performance and predictability of landslide susceptibility mapping methods are influenced by landslide factors. Therefore, landslide factor optimization is a critical task before landslide susceptibility mapping. [ABSTRACT FROM AUTHOR]

Landslide susceptibility consists of an essential component in the day-to-day activity of human beings. Landslide incidents are typically happening at a low rate of recurrence when compared and in contrast to other events. This might be generated into main natural catastrophes relating to widespread and undesirable sound effects. Landslide hotspot area identification and mapping are used for the regional community to secure from this disaster. Therefore, this research aims to identify the hotspot areas of landslide and to generate maps using GIS, AHP, and multi-criteria decision analysis (MCDA). MCDA techniques are applied under such circumstances to categorize and class decisions for successive comprehensive estimation or else to state possible from impossible potentiality with various landslides. Analytical hierarchy process (AHP) constructively applies for conveying influence to different criteria within multi-criteria decision analysis. The causative landslide identifying factors utilized in this research were elevation, slope, aspect, soil type, lithology, distance to stream, land use/land cover, rainfall, and drainage density achieved from various sources. Subsequently, to explain the significance of each constraint into landslide susceptibility, all factors were found using the AHP technique. Generally, landslide susceptibility map factors were multiplied by their weights to acquire with the AHP technique. The result showed that the AHP methods are comparatively good quality estimators of landslide susceptibility identification in the Chemoga watershed. As the result, the Chemoga watershed landslide susceptibility map classes were classified as 46.52%, 13.83%.18.71%, 15.39%, and 5.55% of the occurred landslide fall to very low, low, moderate, high, and very high susceptibility zones, respectively. Performance and accuracy of modeled maps have been established using GPS field data and Google earth data landslide map and area under curve (AUC) of the receiver operating characteristic curve (ROC). As the result, validation depends on the ROC specifies the accuracy of the map formed with the AHP merged through weighted overly method illustrated very good accuracy of AUC value 81.45%. In general, the research outcomes inveterate the very good test consistency of the generated maps. [ABSTRACT FROM AUTHOR]

Mapping the susceptibility of earthquake-induced soil liquefaction at the continental scale is a challenge. Susceptibility of soils to liquefaction is the tendency of certain geomaterials to undergo a severe stiffness degradation and loss of shear strength. The latter could be induced by cyclic loading induced by seismic events. The liquefaction surface evidence is a local phenomenon, and detailed geotechnical field investigations are not available for regional studies, not to mention at continental scales. The literature review shows earthquake-induced soil liquefaction evidence in several European countries, yet, a comprehensive picture of the susceptibility at the European scale is not available. This work aims to develop a methodology to assess the earthquake-induced soil liquefaction susceptibility in Europe using geospatial parameters weighted via the Analytic Hierarchy Process (AHP). The major outcome of the study is a novel Liquefaction Susceptibility map of Europe (LSE), based on the building of ten different European macro-units for earthquake-induced soil liquefaction. These European macro-units have been delineated in this work in order to be homogenous areas from the geological, physiographical, and geomorphological points of view. The adopted input parameters are the depositional environment of the sediments, the distance from water bodies (coast and rivers), and the compound topographic index (as a proxy of the soil saturation). The resolution of the LSE map is 900 × 900 m. The results have been cross-compared with soil liquefaction susceptibility maps available for a region of Greece (i.e. Thrace), Portugal, Bulgaria, and sites where liquefaction manifestations occurred across Europe. The LSE can be adopted to identify at large scale the areas susceptible to liquefaction including also the territories across Europe characterized by low seismicity and potentially affected by anthropogenic seismicity. [ABSTRACT FROM AUTHOR]

Landslides are serious natural disasters frequently occurring in the mountainous areas along the Sichuan–Tibet railway. Landslides are often induced by rainfall and greatly threaten people's lives and property. Via the use of GIS technology, a probabilistic analysis model was proposed for rainfall-induced shallow landslides along the above route. The process of the proposed probabilistic analysis model could be divided into three parts: (i) an infiltration hydrogeological model, (ii) a slope stability model, and (iii) probabilistic analysis. In this model, soil cohesion and internal friction angle were regarded as random parameters. The Green–Ampt model was used to dynamically describe the soil infiltration process during rainfall. Combined with the infinite slope stability model, the failure probability in the study area was analyzed. The model was applied to simulate the Bayi catchment landslide event of August 18, 2010, to evaluate its reliability. Through comparison of the simulation results to landslide occurrence locations, it was determined that the model achieves a satisfactory prediction performance. In addition, compared to deterministic analysis methods and the r.slope model, the proposed probabilistic analysis model achieved a satisfactory evaluation performance. The presented physics-based probabilistic analysis model could provide important theoretical support for hazard prevention in regard to rainfall-induced shallow landslides in mountainous areas. [ABSTRACT FROM AUTHOR]

It is well known that geological structure is one of the main predisposing factors that control landslide occurrence at regional scale. Along time, investigation of the role of structural and lithologic controls has raised the interest of specialists from various earth and planetary sciences who proposed different approaches that range from descriptive analysis to automatic detection. The aim of this study is to investigate the role of structural and lithological controls on landslide occurrence at regional scale. To achieve this, a collection of tools for calculating the TOBIA index and estimating the relationship between morphostructural and lithological conditions and landslide occurrences was developed using Python programming language. The relationship between geological structure and landslide occurrences was analyzed in three distinct landslide-prone sites in Romania, located in different geological and geomorphological contexts. The spatial distribution of the geological structure is modelled based on the TOBIA index (TOpographic/Bedding-plane Intersection Angle). The TOBIA index uses topographic slope, slope aspect, dip angle, and dip direction to divide the terrain into three categories named cataclinal, anaclinal, and orthoclinal, with each category having several subcategories. Additionally, by implementing the bivariate statistical analyses (Weight of Evidence), we investigated the relationship of landslide occurrences to the resulted morphostructural slopes, on the one hand, and to the lithological units, on the other hand. The output contrast values show an evident influence of the cataclinal slopes over landslide occurrences in homoclinal areas, with most of the scarps occurring at the contact between cataclinal and orthoclinal slopes. Fewer influences are found in areas where clay outcrops and where the geological structure has a much-diminished control over the presence of the landslides. These results contribute to a better understanding of the role of structural and lithologic controls on landslide occurrence at regional scale and furthermore could serve to a better assessment of landslide susceptibility. The results obtained on different geologic and geomorphic contexts in Romania suggest that this methodology can be applied successfully in other areas prone to landslides. For reproducibility purposes, the GIS set of tools was made available under the MIT license. [ABSTRACT FROM AUTHOR]

The primary objective of the present research is to apply and compare the performance of evidential belief function (EBF)-based logistic model trees (LMTs) and multiclass alternate decision trees (LADTrees) in landslide susceptibility mapping in Xiaojin County, China. Firstly, 328 landslides were mapped in the study area. Then, 70% of landslide points were used as training samples randomly, and the remaining 30% were intended for validation samples. For the study area, 12 landslide-related conditioning factors were identified, for instance, plan curvature, profile curvature, elevation, slope angle, slope aspect, normalized difference vegetation index (NDVI), topographic wetness index (TWI), land use, lithology, distance to river soil, and distance to roads. The following procedure was to map landslide susceptible regions through EBF, LADTree and LMT models. Finally, the receiver operating characteristic (ROC) curve was utilized to contrast and test the capacity of the three models. The success rates with the training dataset were 0.880, 0.877 and 0.886 for the EBF, LADTree and LMT models, respectively. In addition, their prediction rates with the validation dataset were 0.846, 0.861 and 0.865, respectively. The results could provide references for disaster management and land-use planning. [ABSTRACT FROM AUTHOR]

Landslides have been a great threat to the economy as well as human lives in the north-eastern border state, Manipur, India. Highways serve as the main lifelines of the state where hundreds of goods carrying heavy vehicles are transporting every day. However, during the rainy seasons, these highways are often collapsed due to landslides. This study aims to analyze the effect of multi-resolution Digital Elevation Models (DEMs) on landslide susceptibility mapping along a major highway in Manipur, India. Nine-landslide conditioning factors including Land Use Land Cover (LULC), slope, aspect, curvature, elevation, Normalized Difference Vegetation Index (NDVI), distance to roads, distance to streams, and distance to faults were selected. The topographic factors were extracted from three different DEM sources-CARTOSAT 10 m, ASTER 30 m and SRTM 90 m. The Analytical Hierarchy Process approach is employed to assign appropriate weights to the conditioning factors, and the weighted factors are integrated to produce the landslide susceptibility maps. The assessment of the model accuracies showed that LSM1 (10 m) procured a success accuracy of 83.78% and prediction accuracy of 81.45%, while LSM2 (30 m) resolution procured a success accuracy of 72.8% and prediction accuracy of 70.45%, and LSM3 (90 m) resolution procured a success accuracy of only 58% and prediction accuracy of 55.87%, respectively. In this particular study, the main aim is to analyse the effect of resolution in Landslide Susceptibility mapping along a major highway in Manipur, India, which has been witnessing landslides every year. The study also aims to select the optimal DEM resolution. The selection for optimal DEM resolution is dependent on scale and the topographic setting of the region under consideration. The susceptibility mapping results of this study indicates that CARTOSAT 10 m resolution DEM produced better accuracy as compared to other DEMs for this particular site. Policy makers can use the resulting map in planning purposes and risk management. [ABSTRACT FROM AUTHOR]

Landslide hazard assessment is critical for preventing and mitigating landslide disasters. The tuning of hyperparameters is of great importance to achieve better accuracy in a landslide hazard assessment model. In this study, a novel approach is proposed for landslide hazard assessment with support vector machine (SVM) as the primary model and Bayesian optimization (BO) algorithm as the parameter tuning method. This study describes 1711 historical landslide disaster points in Nanping City, and a total of 12 landslide conditioning factors including elevation, slope, aspect, curvature, lithology, soil type, soil erosion, rainfall, river, land use, highway, and railway were selected. The multicollinearity diagnosis was performed on the factors using the Spearman correlation coefficient. For model validation, 1711 landslides and 1711 non-landslides were collected as the dataset and divided into a training dataset (50 %) and a testing dataset (50 %). The performance of the model was evaluated by the confusion matrix and receiver operating characteristic (ROC) curve. The results of the confusion matrix accuracy and the area under the ROC curve showed that the BO-SVM model (89.53 %, 0.97) performed better than the SVM model (84.91 %, 0.93). In addition, the landslide hazard maps generated by the BO-SVM model had better overall results than that by the SVM model. [ABSTRACT FROM AUTHOR]