Comparative study of different machine learning models in landslide susceptibility assessment: A case study of Conghua District, Guangzhou, China

Machine learning is currently one of the research hotspots in the field of landslide prediction.To clarify and evaluate the differences in characteristics and prediction effects of different machine learning models,Conghua District,which is the most prone to landslide disasters in Guangzhou,was selected for landslide susceptibility evaluation.The evaluation factors were selected by using correlation analysis and variance expansion factor method.Applying four machine learning methods namely Logistic Regression(LR),Random Forest(RF),Support Vector Machines(SVM),and Extreme Gradient Boosting(XGB),landslide models were constructed.Comparative analysis and evaluation of the model were conducted through statistical indices and receiver operating characteristic(ROC)curves.The results showed that LR,RF,SVM,and XGB models have good predictive performance for landslide susceptibility,with the area under curve(AUC)values of 0.752,0.965,0.996,and 0.998,respectively.XGB model had the highest predictive ability,followed by RF model,SVM model,and LR model.The frequency ratio(FR)accuracy of LR,RF,SVM,and XGB models was 0.775,0.842,0.759,and 0.822,respectively.RF and XGB models were superior to LR and SVM models,indicating that the integrated algorithm has better predictive ability than a single classification algorithm in regional landslide classification problems.

Comparison of LR,5-CV SVM,GA SVM,and PSO SVM for landslide susceptibility assessment in Tibetan Plateau area,China

The applicability of statistics-based landslide susceptibility assessment methods is affected by the number of historical landslides.Previous studies have proposed support vector machine(SVM)as a small-sample learning method.However,those studies demonstrated that different parameters can affect model performance.We optimized the SVM and obtained models as 5-fold cross validation(5-CV)SVM,genetic algorithm(GA)SVM,and particle swarm optimization(PSO)SVM.This study compared the prediction performances of logistic regression(LR),5-CV SVM,GA SVM,and PSO SVM on landslide susceptibility mapping,to explore the spatial distribution of landslide susceptibility in the study area in Tibetan Plateau,China.A geospatial database was established based on 392 historical landslides and 392 non-landslides in the study area.We used 11 influencing factors of altitude,slope,aspect,curvature,lithology,normalized difference vegetation index(NDVI),distance to road,distance to river,distance to fault,peak ground acceleration(PGA),and rainfall to construct an influencing factor evaluation system.To evaluate the models,four susceptibility maps were compared via receiver operating characteristics(ROC)curve and the results showed that prediction rates for the models are 84%(LR),87%(5-CV SVM),85%(GA SVM),and 90%(PSO SVM).We also used precision,recall,F1-score and accuracy to assess the quality performance of these models.The results showed that the PSO SVM had greater potential for future implementation in the Tibetan Plateau area because of its superior performance in the landslide susceptibility assessment.

Effects of the probability of pulse-like ground motions on landslide susceptibility assessment in near-fault areas

Earthquake-induced strong near-fault ground motion is typically accompanied by largevelocity pulse-like component,which causes serious damage to slopes and buildings.Although not all near-fault ground motions contain a pulse-like component,it is important to consider this factor in regional earthquake-induced landslide susceptibility assessment.In the present study,we considered the probability of the observed pulse-like ground motion at each site(PP)in the region of an earthquake as one of the conditioning factors for landslide susceptibility assessment.A subset of the area affected by the 1994Mw6.7 Northridge earthquake in California was examined.To explore and verify the effects of PP on landslide susceptibility assessment,seven models were established,consisting of six identical influencing factors(elevation,slope gradient,aspect,distance to drainage,distance to roads,and geology)and one or two factors characterizing the intensity of the earthquake(distance to fault,peak ground acceleration,peak ground velocity,and PP)in logistic regression analysis.The results showed that the model considering PP performed better in susceptibility assessment,with an area under the receiver operating characteristic curve value of 0.956.Based on the results of relative importance analysis,the contribution of the PP value to earthquakeinduced landslide susceptibility was ranked fourth after the slope gradient,elevation,and lithology.The prediction performance of the model considering the pulse-like effect was better than that reported previously.A logistic regression model that considers the pulse-like effect can be applied in disaster prevention,mitigation,and construction planning in near-fault areas.

Landslide susceptibility assessment using the certainty factor and analytic hierarchy process

A new approach combining the certainty factor(CF) and analytic hierarchy process(AHP) methods was proposed to assess landslide susceptibility in the Ziyang district, which is situated in the Qin-Ba Mountain region, China. Landslide inventory data were collected based on field investigations and remote sensing interpretations. A total of 791 landslides were identified. A total of 633 landslides were randomly selected from this data setas the training set, and the remaining landslides were used for validation as the test set. Nine factors, including the slope angle, slope aspect, slope curvature, lithology, distance to faults, distance to streams, precipitation, road network intensity degree and land use were chosen as the landslide causal factors for further susceptibility assessment. The weight of each factor and its subclass were calculated by AHP and CF methods. Landslide susceptibility was compared between the bivariate statistical method and the proposed CF-AHP method. The results indicate that the distance to streams, distance to faults and lithology are the most dominant causal factors associated with landslides. The susceptibility zonation was categorized into five classes of landslide susceptibility, i.e., very high, high, moderate, low and very low level. Lastly, the relative operating characteristics(ROC) curve was used to validate the accuracy of the new approach, and the result showed a satisfactory prediction rate of 78.3%, compared to 69.2% obtained with the landslide susceptibility index method. The results indicate that the CF-AHP combined method is more appropriate for assessing the landslide susceptibility in this area.

Landslide inventory and susceptibility assessment using multiple statistical approaches along the Karakoram highway,northern Pakistan

China-Pakistan Economic Corridor(CPEC)is a framework of regional connectivity,which will not only benefit China and Pakistan but will have positive impact on Iran,Afghanistan,India,Central Asian Republic,and the region.The surrounding area in CPEC is prone to frequent disruption by geological hazards mainly landslides in northern Pakistan.Comprehensive landslide inventory and susceptibility assessment are rarely available to utilize for landslide mitigation strategies.This study aims to utilize the high-resolution satellite images to develop a comprehensive landslide inventory and subsequently develop landslide susceptibility maps using multiple techniques.The very high-resolution(VHR)satellite images are utilized to develop a landslide inventory using the visual image classification techniques,historic records and field observations.A total of 1632 landslides are mapped in the area.Four statistical models i.e.,frequency ratio,artificial neural network,weights of evidence and logistic regression were used for landslide susceptibility modeling by comparing the landslide inventory with the topographic parameters,geological features,drainage and road network.The developed landslides susceptibility maps were verified using the area under curve(AUC)method.The prediction power of the model was assessed by the prediction rate curve.The success rate curves show 93%,92.8%,92.7%and 87.4%accuracy of susceptibility maps for frequency ratio,artificial neural network,weights of evidence and logistic regression,respectively.The developed landslide inventory and susceptibility maps can be used for land use planning and landslide mitigation strategies.

Optimizing the frequency ratio method for landslide susceptibility assessment: A case study of the Caiyuan Basin in the southeast mountainous area of China

Bivariate statistical analysis of data-driven approaches is widely used for landslide susceptibility assessment, and the frequency ratio(FR) method is one of the most popular. However, the results of such assessments are dominated by the number of classes and bounds of landslide-related causative factors, and the optimal assessment is unknown. This paper optimizes the frequency ratio method as an example of bivariate statistical analysis for landslide susceptibility mapping based on a case study of the Caiyuan Basin, a region with frequent landslides, which is located in the southeast coastal mountainous area of China. A landslide inventory map containing a total of 1425 landslides(polygons) was produced, in which 70% of the landslides were selected for training purposes, and the remaining were used for validationpurposes. All datasets were resampled to the same 5 m × 5 m/pixel resolution. The receiver operating characteristic(ROC) curves of the susceptibility maps were obtained based on different combinations of dominating parameters, and the maximum value of the areas under the ROC curves(AUCs) as well as the corresponding optimal parameter was identified with an automatic searching algorithm. The results showed that the landslide susceptibility maps obtained using optimal parameters displayed a significant increase in the prediction AUC compared with those values obtained using stochastic parameters. The results also showed that one parameter named bin width has a dominant influence on the optimum. In practice, this paper is expected to benefit the assessment of landslide susceptibility by providing an easy-to-use tool. The proposed automatic approach provides a way to optimize the frequency ratio method or other bivariate statistical methods, which can furtherfacilitate comparisons and choices between different methods for landslide susceptibility assessment.

A Hybrid Multi-Hazard Susceptibility Assessment Model for a Basin in Elazig Province,Türkiye

Preparation of accurate and up-to-date susceptibility maps at the regional scale is mandatory for disaster mitigation,site selection,and planning in areas prone to multiple natural hazards.In this study,we proposed a novel multi-hazard susceptibility assessment approach that combines expert-based and supervised machine learning methods for landslide,flood,and earthquake hazard assessments for a basin in Elazig Province,Türkiye.To produce the landslide susceptibility map,an ensemble machine learning algorithm,random forest,was chosen because of its known performance in similar studies.The modified analytical hierarchical process method was used to produce the flood susceptibility map by using factor scores that were defined specifically for the area in the study.The seismic hazard was assessed using ground motion parameters based on Arias intensity values.The univariate maps were synthesized with a Mamdani fuzzy inference system using membership functions designated by expert.The results show that the random forest provided an overall accuracy of 92.3%for landslide susceptibility mapping.Of the study area,41.24%were found prone to multi-hazards(probability value>50%),but the southern parts of the study area are more susceptible.The proposed model is suitable for multi-hazard susceptibility assessment at a regional scale although expert intervention may be required for optimizing the algorithms.

Identification of landslide spatial distribution and susceptibility assessment in relation to topography in the Xi＇an Region, Shaanxi Province, China

Landslides are among the most serious of geohazards in the Xi＇an Region, Shaanxi, China, and are responsible for extensive human and property loss. In order to understand the distribution of landslides and assess their associated hazards in this region, we used a combination of frequency analysis, logistic analysis, and Geographic Information System （GIS） analysis, with consideration of the spatial distribution of landslides. Using the GIS approach, the five key factors of surface topography, including slope gradient, topographic wetness index （TWI）, height difference, profile curvature and slope aspect, were considered. First, the distribution and frequency of landslides were considered in relation to all of the five factors in each of three sub-regions susceptible to landslides （Qin Mountain, Li Mountain, and Loess Tableland）. Secondly, each factor＇s influence was deter- mined by a logistic regression method, and the relative importance of each of these independent variables was evaluated. Finally, a landslide susceptibility map was generated using GIS tools. Locations that had recorded landslides were used to validate the results of the landslide susceptibility map and the accuracy obtained was above 84%. The validation proved that there is sufficient agreement between the susceptibility map and existing records of landslide occurrences. The logistic regression model produced acceptable results （the areas under the Receiver Operating Characteristics （ROC） curve were 0.865, 0.841, and 0.924 in the Qin Mountain, Li Mountain and Loess Tableland）. We are confident that the results of this study can be useful in preliminary planning for land use, particularly for construction work in high-risk areas.

Probabilistic Study of Landslide Susceptibility in the Sagrado River Watershed,Brazil

In this paper,we present a probabilistic study of landslide susceptibility at the Sagrado River Watershed in Morretes,Brazil.The area is characterized by slopes higher than 20%in a large part of the relief,variable soil depths,and by strong rainfall intensities due to orographic rains.Taken together,these factors promote the occurrence of translational landslides.Anthropic occupation is distributed along the lowlands and on the less inclined slopes of the Serra do Mar.The infinite slope method was used to determine the distribution of susceptibility to landslides.Input parameters of the model consisted of geotechnical soil parameters,soil layer thickness,slope of the hillside,and matric suction,and the analysis implemented a Monte Carlo probabilistic method.As such,this method allows the quantification of uncertainties due to the variability in geotechnical parameters,which enables the determination of a probability of rupture.The elaboration of susceptibility maps to landslides for unsaturated soil conditions represents a useful tool to support the identification of critical events occurrences in this location.

Analysing post-earthquake landslide susceptibility using multi-temporal landslide inventories——a case study in Miansi Town of China

The earthquake that occurred on May 12, 2008, in Wenchuan County aroused a great deal of research on co-seismic landslide susceptibility assessment, but there is still a lack of an evaluation method that considers the activity state of the landslide itself. Therefore, this paper establishes a new susceptibility evaluation model that superimposes the active landslide state based on previous susceptibility evaluation models. Based on a multi-phase landslide database, the probabilistic approach was used to evaluate landslide susceptibility in the Miansi town over many years. We chose the elevation, slope, aspect, and distance from the channel as trigger factors and then used the probability comprehensive discrimination method to calculate the probability of landslide occurrence. Then, the susceptibility results of each period were calculated by superposition with the activity rate. The results show that between 2008 and 2014, the proportion of areas with low landslide susceptibility in the study area was the largest, and the proportionof areas with the highest susceptibility was minimal. The landslide area with highest susceptibility gradually decreased from 2014 to 2017. However, in 2017, 15.06% of the area was still with high susceptibility, and relevant disaster prevention and reduction measures should be taken in these areas. The larger area under the receiver operating characteristic curve(AUC) indicates that the results of the landslide susceptibility assessment in this study are more objective and reliable than those of previous models. The difference in the AUC values over many years shows that the accuracy of the evaluation results of this model is not constant, and a greater number of landslides or higher landslide activity corresponds to a higher accuracy of the evaluation results.

Reliable assessment approach of landslide susceptibility in broad areas based on optimal slope units and negative samples involving priori knowledge

Reliable assessment of landslide susceptibility in broad areas of terrain remains challenging due to complex topography and poor representation of randomly selected negative samples.Assessment in broad areas is now primarily based on grid units,which do not have a clear physical meaning like slope units,and their accuracy is not ideal.Nevertheless,the large amount of manual editing,due to the incorrectly generated horizontal and vertical lines during slope unit partitioning,limits using slope units for rapid assessment over large areas.Hence,this paper proposes a reliable susceptibility assessment approach to solve this problem based on optimal slope units and negative samples involving prior knowledge.Precisely,an algorithm to automatically extract slope units is designed to eliminate fragmented and erroneous units.Second,a samples labeling index(SLI)is defined based on the certainty factors model to select negative samples reasonably.Sichuan Province,China is selected for experimental analysis,with the results demonstrate that the optimized slope unit and the negative samples selection strategy consider prior knowledge achieve better results in the random forest model,support vector machine model,and artificial neural network model.In particular,the composite performance index AUC of artificial neural network model improved from 0.81 to 0.90.