Landslide susceptibility mapping using GIS-based multi-criteria decision analysis,support vector machines,and logistic regression

Identification of landslides and production of landslide susceptibility maps are crucial steps that can help planners, local administrations, and decision makers in disaster planning. Accuracy of the landslide susceptibility maps is important for reducing the losses of life and property. Models used for landslide susceptibility mapping require a combination of various factors describing features of the terrain and meteorological conditions. Many algorithms have been developed and applied in the literature to increase the accuracy of landslide susceptibility maps. In recent years, geographic information system-based multi-criteria decision analyses (MCDA) and support vector regression (SVR) have been successfully applied in the production of landslide susceptibility maps. In this study, the MCDA and SVR methods were employed to assess the shallow landslide susceptibility of Trabzon province (NE Turkey) using lithology, slope, land cover, aspect, topographic wetness index, drainage density, slope length, elevation, and distance to road as input data. Performances of the methods were compared with that of widely used logistic regression model using ROC and success rate curves. Results showed that the MCDA and SVR outperformed the conventional logistic regression method in the mapping of shallow landslides. Therefore, multi-criteria decision method and support vector regression were employed to determine potential landslide zones in the study area.     

A comparative study on the landslide susceptibility mapping using logistic regression and statistical index models

The logistic regression and statistical index models are applied and verified for landslide susceptibility mapping in Daguan County, Yunnan Province, China, by means of the geographic information system (GIS). A detailed landslide inventory map was prepared by literatures, aerial photographs, and supported by field works. Fifteen landslide-conditioning factors were considered: slope angle, slope aspect, curvature, plan curvature, profile curvature, altitude, STI, SPI, and TWI were derived from digital elevation model; NDVI was extracted from Landsat ETM7; rainfall was obtained from local rainfall data; distance to faults, distance to roads, and distance to rivers were created from a 1:25,000 scale topographic map; the lithology was extracted from geological map. Using these factors, the landslide susceptibility maps were prepared by LR and SI models. The accuracy of the results was verified by using existing landslide locations. The statistical index model had a predictive rate of 81.02%, which is more accurate prediction in comparison with logistic regression model (80.29%). The models can be used to land-use planning in the study area.     

A case study using support vector machines, neural networks and logistic regression in a GIS to identify wells contaminated with nitrate-N

Accurate and inexpensive identification of potentially contaminated wells is critical for water resources protection and management. The objectives of this study are to 1) assess the suitability of approximation tools such as neural networks (NN) and support vector machines (SVM) integrated in a geographic information system (GIS) for identifying contaminated wells and 2) use logistic regression and feature selection methods to identify significant variables for transporting contaminants in and through the soil profile to the groundwater. Fourteen GIS derived soil hydrogeologic and landuse parameters were used as initial inputs in this study. Well water quality data (nitrate-N) from 6,917 wells provided by Florida Department of Environmental Protection (USA) were used as an output target class. The use of the logistic regression and feature selection methods reduced the number of input variables to nine. Receiver operating characteristics (ROC) curves were used for evaluation of these approximation tools. Results showed superior performance with the NN as compared to SVM especially on training data while testing results were comparable. Feature selection did not improve accuracy; however, it helped increase the sensitivity or true positive rate (TPR). Thus, a higher TPR was obtainable with fewer variables.     

Spatial prediction models for shallow landslide hazards: a comparative assessment of the efficacy of support vector machines,artificial neural networks,kernel logistic regression,and logistic model tree

Preparation of landslide susceptibility maps is considered as the first important step in landslide risk assessments, but these maps are accepted as an end product that can be used for land use planning. The main objective of this study is to explore some new state-of-the-art sophisticated machine learning techniques and introduce a framework for training and validation of shallow landslide susceptibility models by using the latest statistical methods. The Son La hydropower basin (Vietnam) was selected as a case study. First, a landslide inventory map was constructed using the historical landslide locations from two national projects in Vietnam. A total of 12 landslide conditioning factors were then constructed from various data sources. Landslide locations were randomly split into a ratio of 70:30 for training and validating the models. To choose the best subset of conditioning factors, predictive ability of the factors were assessed using the Information Gain Ratio with 10-fold cross-validation technique. Factors with null predictive ability were removed to optimize the models. Subsequently, five landslide models were built using support vector machines (SVM), multi-layer perceptron neural networks (MLP Neural Nets), radial basis function neural networks (RBF Neural Nets), kernel logistic regression (KLR), and logistic model trees (LMT). The resulting models were validated and compared using the receive operating characteristic (ROC), Kappa index, and several statistical evaluation measures. Additionally, Friedman and Wilcoxon signed-rank tests were applied to confirm significant statistical differences among the five machine learning models employed in this study. Overall, the MLP Neural Nets model has the highest prediction capability (90.2 %), followed by the SVM model (88.7 %) and the KLR model (87.9 %), the RBF Neural Nets model (87.1 %), and the LMT model (86.1 %). Results revealed that both the KLR and the LMT models showed promising methods for shallow landslide susceptibility mapping. The result from this study demonstrates the benefit of selecting the optimal machine learning techniques with proper conditioning selection method in shallow landslide susceptibility mapping.     

A comparative study of landslide susceptibility mapping using weight of evidence,logistic regression and support vector machine and evaluated by SBAS-InSAR monitoring: Zhouqu to Wudu segment in Bailong River Basin,China

The determining of landslide-prone areas in mountainous terrain is essential for land planning and hazard mitigation. In this paper, a comparative study using three statistical models including weight of evidence model (WoE), logistic regression model (LR) and support vector machine method (SVM) was undertaken in the Zhouqu to Wudu segment in the Bailong River Basin, Southern Gansu, China. Six conditionally independent environmental factors, elevation, slope, aspect, distance from fault, lithology and settlement density, were selected as the explanatory variables that may contribute to landslide occurrence based on principal component analysis (PCA) and Chi-square test. The relation between landslide distributions and these variables was analyzed using the three models and the results then used to calculate the landslide susceptibility (LS). The performance of the models was then evaluated using both the highly accurate deformation signals produced by using the Small Baseline Subset Interferometric Synthetic Aperture Radar technique and Receiver Operating Characteristic (ROC) curve. Results show more deformation points in areas with high and very high LS levels, and also more stable points in areas with low and very low LS levels for the SVM model. In addition, the SVM has larger area under the ROC curve. It indicates that the SVM has better prediction accuracy and classified ability. For the interpretability, the WoE derives the class of factors that most contributed to landsliding in the study area, and the LR reveals that factors including elevation, settlement density and distance from fault played major roles in landslide occurrence and distribution, whereas the SVM cannot provide relative weights for the variables. The outperformed SVM could be employed to determine potential landslide zones in the study area. Outcome of this research would provide preliminary basis for general land planning such as choosing new urban areas and infrastructure construction in the future, as well as for landslide hazard mitigation in Bailong River Basin.     

Modelling shallow landslide susceptibility: a new approach in logistic regression by using favourability assessment

A new method for estimating shallow landslide susceptibility by combining Geographical Information System (GIS), nonparametric kernel density estimation and logistic regression is described. Specifically, a logistic regression is applied to predict the spatial distribution by estimating the probability of occurrence of a landslide in a 16 km² area. For this purpose, a GIS is employed to gather the relevant sample information connected with the landslides. The advantages of pre-processing the explanatory variables by nonparametric density estimation (for continuous variables) and a reclassification (for categorical/discrete ones) are discussed. The pre-processing leads to new explanatory variables, namely, some functions which measure the favourability of occurrence of a landslide. The resulting model correctly classifies 98.55% of the inventaried landslides and 89.80% of the landscape surface without instabilities. New data about recent shallow landslides were collected in order to validate the model, and 92.20% of them are also correctly classified. The results support the methodology and the extrapolation of the model to the whole study area (278 km²) in order to obtain susceptibility maps.     

Mapping landslide susceptibility with logistic regression, multiple adaptive regression splines, classification and regression trees, and maximum entropy methods: a comparative study

Four statistical techniques for modelling landslide susceptibility were compared: multiple logistic regression (MLR), multivariate adaptive regression splines (MARS), classification and regression trees (CART), and maximum entropy (MAXENT). According to the literature, MARS and MAXENT have never been used in landslide susceptibility modelling, and CART has been used only twice. Twenty independent variables were used as predictors, including lithology as a categorical variable. Two sets of random samples were used, for a total of 90 model replicates (with and without lithology, and with different proportions of positive and negative data). The model performance was evaluated using the area under the receiver operating characteristic curve (AUC) statistic. The main results are (a) the inclusion of lithology improves the model performance; (b) the best AUC values for single models are MLR (0.76), MARS (0.76), CART (0.77), and MAXENT (0.78); (c) a smaller amount of negative data provides better results; (d) the models with the highest prediction capability are obtained with MAXENT and CART; and (e) the combination of different models is a way to evaluate the model reliability. We further discuss some key issues in landslide modelling, including the influence of the various methods that we used, the sample size, and the random replicate procedures.     

A novel hybrid integration model using support vector machines and random subspace for weather-triggered landslide susceptibility assessment in the Wuning area (China)

This study proposed a hybrid modeling approach using two methods, support vector machines and random subspace, to create a novel model named random subspace-based support vector machines (RSSVM) for assessing landslide susceptibility. The newly developed model was then tested in the Wuning area, China, to produce a landslide susceptibility map. With the purpose of achieving the objective of the study, a spatial dataset was initially constructed that includes a landslide inventory map consisting of 445 landslide regions. Then, various landslide-influencing factors were defined, including slope angle, aspect, altitude, topographic wetness index, stream power index, sediment transport index, soil, lithology, normalized difference vegetation index, land use, rainfall, distance to roads, distance to rivers, and distance to faults. Next, the result of the RSSVM model was validated using statistical index-based evaluations and the receiver operating characteristic curve approach. Then, to evaluate the performance of the suggested RSSVM model, a comparison analysis was performed to other existing approaches such as artificial neural network, Naïve Bayes (NB) and support vector machine (SVM). In general, the performance of the RSSVM model was better than the other models for spatial prediction of landslide susceptibility. The AUC results of the applied models are as follows: RSSVM (AUC = 0.857), followed by MLP (AUC = 0.823), SVM (AUC = 0.814) and NB (AUC = 0.783). The present study indicates that RSSVM can be used for landslide susceptibility evaluation, and the results are very useful for local governments and people living in the Wuning area.     

Combining landslide susceptibility maps obtained from frequency ratio,logistic regression,and artificial neural network models using ASTER images and GIS

Landslide-related factors were extracted from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images, and integrated techniques were developed, applied, and verified for the analysis of landslide susceptibility in Boun, Korea, using a geographic information system (GIS). Digital elevation model (DEM), lineament, normalized difference vegetation index (NDVI), and land-cover factors were extracted from the ASTER images for analysis. Slope, aspect, and curvature were calculated from a DEM topographic database. Using the constructed spatial database, the relationships between the detected landslide locations and six related factors were identified and quantified using frequency ratio (FR), logistic regression (LR), and artificial neural network (ANN) models. These relationships were used as factor ratings in an overlay analysis to create landslide susceptibility indices and maps. Three landslide susceptibility maps were then combined and applied as new input factors in the FR, LR, and ANN models to make improved susceptibility maps. All of the susceptibility maps were verified by comparison with known landslide locations not used for training the models. The combined landslide susceptibility maps created using three landslide-related input factors showed improved accuracy (87.00% in FR, 88.21% in LR, and 86.51% in ANN models) compared to the individual landslide susceptibility maps (84.34% in FR, 85.40% in LR, and 74.29% in ANN models) generated using the six factors from the ASTER images.     

Comparison of landslide susceptibility mapping methodologies for Koyulhisar,Turkey: conditional probability,logistic regression,artificial neural networks,and support vector machine

This case study presented herein compares the GIS-based landslide susceptibility mapping methods such as conditional probability (CP), logistic regression (LR), artificial neural networks (ANNs) and support vector machine (SVM) applied in Koyulhisar (Sivas, Turkey). Digital elevation model was first constructed using GIS software. Landslide-related factors such as geology, faults, drainage system, topographical elevation, slope angle, slope aspect, topographic wetness index, stream power index, normalized difference vegetation index, distance from settlements and roads were used in the landslide susceptibility analyses. In the last stage of the analyses, landslide susceptibility maps were produced from ANN, CP, LR, SVM models, and they were then compared by means of their validations. However, area under curve values obtained from all four methodologies showed that the map obtained from ANN model looks like more accurate than the other models, accuracies of all models can be evaluated relatively similar. The results also showed that the CP is a simple method in landslide susceptibility mapping and highly compatible with GIS operating features. Susceptibility maps can be easily produced using CP, because input process, calculation and output processes are very simple in CP model when compared with the other methods considered in this study.     

Application of logistic regression and fuzzy operators to landslide susceptibility assessment in Azdavay (Kastamonu,Turkey)

Landslides and their assessments are of great importance since they damage properties, infrastructures, environment, lives and so on. Particularly, landslide inventory, susceptibility, and hazard or risk mapping have become important issues in the last few decades. Such maps provide useful information and can be produced by qualitative or quantitative methods. The work presented in this paper aimed to assess landslide susceptibility in a selected area, covering 570.625 km² in the Western Black Sea region of Turkey, by two quantitative methods. For this purpose, in the first stage, a detailed landslide inventory map was prepared by extensive field studies. A total of 96 landslides were mapped during these studies. To perform landslide susceptibility analyses, six input parameters such as topographical elevation, lithology, land use, slope, aspect and distance to streams were considered. Two quantitative methods, logistic regression and fuzzy approach, were used to assess landslide susceptibility in the selected area. For the fuzzy approach, the fuzzy and, or, algebraic product, algebraic sum and gamma operators were considered. At the final stage, 18 landslide susceptibility maps were produced by the logistic regression and fuzzy operators in a GIS (Geographic Information System) environment. Two performance indicators such as ROC (relative operating characteristics) and cosine amplitude method (r _ij ) were used to validate the final susceptibility maps. Based on the analyses, the landslide susceptibility map produced by the fuzzy gamma operator with a level of 0.975 showed the best performance. In addition, the maps produced by the logistic regression, fuzzy algebraic product and the higher levels of gamma operators showed more satisfactory results, while the fuzzy and, or, algebraic sum maps were not sufficient to provide reliable outputs.     

Forward logistic regression for earth-flow landslide susceptibility assessment in the Platani river basin (southern Sicily,Italy)

Forward logistic regression has allowed us to derive an earth-flow susceptibility model for the Tumarrano river basin, which was defined by modeling the statistical relationships between an archive of 760 events and a set of 20 predictors. For each landslide in the inventory, a landslide identification point (LIP) was automatically produced as corresponding to the highest point along the boundary of the landslide polygons, and unstable conditions were assigned to cells at a distance up to 8 m. An equal number of stable cells (out of landslides) was then randomly extracted and appended to the LIPs to prepare the dataset for logistic regression. A model building strategy was applied to enlarge the area included in training the model and to verify the sensitivity of the regressed models with respect to the locations of the selected stable cells. A suite of 16 models was prepared by randomly extracting different unoverlapping stable cell subsets that have been appended to the unstable ones. Models were finally submitted to forward logistic regression and validated. The results showed satisfying and stable error rates (0.236 on average, with a standard deviation of 0.007) and areas under the receiver operating characteristic (ROC) curve (AUCs) (0.839 for training and 0.817 for test datasets) as well as factor selections (ranks and coefficients). As regards the predictors, steepness and large-profile and local-plan topographic curvatures were systematically selected. Clayey outcropping lithology, midslope drainage, local and midslope ridges, and canyon landforms were also very frequently (from eight to 15 times) included in the models by the forward selection procedures. The model-building strategy allowed us to produce a performing earth-flow susceptibility model, whose model fitting, prediction skill, and robustness were estimated on the basis of validation procedures, demonstrating the independence of the regressed model on the specific selection of the stable cells.     

Spatial prediction of landslide susceptibility in western Serbia using hybrid support vector regression (SVR) with GWO,BAT and COA algorithms

In this study, we developed multiple hybrid machine-learning models to address parameter optimization limitations and enhance the spatial prediction of landslide susceptibility models. We created a geographic information system database, and our analysis results were used to prepare a landslide inventory map containing 359 landslide events identified from Google Earth, aerial photographs, and other validated sources. A support vector regression (SVR) machine-learning model was used to divide the landslide inventory into training (70%) and testing (30%) datasets. The landslide susceptibility map was produced using 14 causative factors. We applied the established gray wolf optimization (GWO) algorithm, bat algorithm (BA), and cuckoo optimization algorithm (COA) to fine-tune the parameters of the SVR model to improve its predictive accuracy. The resultant hybrid models, SVR-GWO, SVR-BA, and SVR-COA, were validated in terms of the area under curve (AUC) and root mean square error (RMSE). The AUC values for the SVR-GWO (0.733), SVR-BA (0.724), and SVR-COA (0.738) models indicate their good prediction rates for landslide susceptibility modeling. SVR-COA had the greatest accuracy, with an RMSE of 0.21687, and SVR-BA had the least accuracy, with an RMSE of 0.23046. The three optimized hybrid models outperformed the SVR model (AUC = 0.704, RMSE = 0.26689), confirming the ability of metaheuristic algorithms to improve model performance.     

Geomorphology and landslide susceptibility assessment using GIS and bivariate statistics: a case study in southern Italy

In this article, the results of a study aimed to assess the landslide susceptibility in the Calaggio Torrent basin (Campanian Apennines, southern Italy) are presented. The landslide susceptibility has been assessed using two bivariate-statistics-based methods in a GIS environment. In the first method, widely used in the existing literature, weighting values (Wi) have been calculated for each class of the selected causal factors (lithology, land-use, slope angle and aspect) taking into account the landslide density (detachment zones + landslide body) within each class. In the second method, which is a modification of the first method, only the landslide detachment zone (LDZ) density has been taken into account to calculate the weighting values. This latter method is probably characterized by a major geomorphological coherence. In fact, differently from the landslide bodies, LDZ must necessarily occur in geoenvironmental classes prone to failure. Thus, the calculated Wi seem to be more reliable in estimating the propensity of a given class to generate failure. The thematic maps have been reclassified on the basis of the calculated Wi and then overlaid, with the purpose to produce landslide susceptibility maps. The used methods converge both in indicating that most part of the study area is characterized by a high–very high landslide susceptibility and in the location and extent of the low-susceptible areas. However, an increase of both the high–very high and moderate–high susceptible areas occurs in using the second method. Both the produced susceptibility maps have been compared with the geomorphological map, highlighting an excellent coherence which is higher using method-2. In both methods, the percentage of each susceptibility class affected by landslides increases with the degree of susceptibility, as expected. However, the percentage at issue in the lowest susceptibility class obtained using method-2, even if low, is higher than that obtained using method-1. This suggests that method-2, notwithstanding its major geomorphological coherence, probably still needs further refinements.     

Application of logistic regression for landslide susceptibility zoning of Cekmece Area, Istanbul, Turkey

As a result of industrialization, throughout the world, cities have been growing rapidly for the last century. One typical example of these growing cities is Istanbul, the population of which is over 10 million. Due to rapid urbanization, new areas suitable for settlement and engineering structures are necessary. The Cekmece area located west of the Istanbul metropolitan area is studied, because the landslide activity is extensive in this area. The purpose of this study is to develop a model that can be used to characterize landslide susceptibility in map form using logistic regression analysis of an extensive landslide database. A database of landslide activity was constructed using both aerial-photography and field studies. About 19.2% of the selected study area is covered by deep-seated landslides. The landslides that occur in the area are primarily located in sandstones with interbedded permeable and impermeable layers such as claystone, siltstone and mudstone. About 31.95% of the total landslide area is located at this unit. To apply logistic regression analyses, a data matrix including 37 variables was constructed. The variables used in the forwards stepwise analyses are different measures of slope, aspect, elevation, stream power index (SPI), plan curvature, profile curvature, geology, geomorphology and relative permeability of lithological units. A total of 25 variables were identified as exerting strong influence on landslide occurrence, and included by the logistic regression equation. Wald statistics values indicate that lithology, SPI and slope are more important than the other parameters in the equation. Beta coefficients of the 25 variables included the logistic regression equation provide a model for landslide susceptibility in the Cekmece area. This model is used to generate a landslide susceptibility map that correctly classified 83.8% of the landslide-prone areas.     

Using multiple logistic regression and GIS technology to predict landslide hazard in northeast Kansas, USA

Landslides in the hilly terrain along the Kansas and Missouri rivers in northeastern Kansas have caused millions of dollars in property damage during the last decade. To address this problem, a statistical method called multiple logistic regression has been used to create a landslide-hazard map for Atchison, Kansas, and surrounding areas. Data included digitized geology, slopes, and landslides, manipulated using ArcView GIS. Logistic regression relates predictor variables to the occurrence or nonoccurrence of landslides within geographic cells and uses the relationship to produce a map showing the probability of future landslides, given local slopes and geologic units. Results indicated that slope is the most important variable for estimating landslide hazard in the study area. Geologic units consisting mostly of shale, siltstone, and sandstone were most susceptible to landslides. Soil type and aspect ratio were considered but excluded from the final analysis because these variables did not significantly add to the predictive power of the logistic regression. Soil types were highly correlated with the geologic units, and no significant relationships existed between landslides and slope aspect.