Landslide Susceptibility Mapping Based on Random Forest and Boosted Regression Tree Models, and a Comparison of Their Performance

Abstract

This study aims to analyze and compare landslide susceptibility at Woomyeon Mountain, South Korea, based on the random forest (RF) model and the boosted regression tree (BRT) model. Through the construction of a landslide inventory map, 140 landslide locations were found. Among these, 42 (30%) were reserved to validate the model after 98 (70%) had been selected at random for model training. Fourteen landslide explanatory variables related to topography, hydrology, and forestry factors were considered and selected, based on the results of information gain for the modeling. The results were evaluated and compared using the receiver operating characteristic curve and statistical indices. The analysis showed that the RF model was better than the BRT model. The RF model yielded higher specificity, overall accuracy, and kappa index than the BRT model. In addition, the RF model, with a prediction rate of 0.865, performed slightly better than the BRT model, which had a prediction rate of 0.851. These results indicate that the landslide susceptibility maps (LSMs) produced in this study had good performance for predicting the spatial landslide distribution in the study area. These LSMs could be helpful for establishing mitigation strategies and for land use planning. deforestation, increased urbanization, and an increase in regional precipitation in landslide-prone areas due to climate change [5]. It is essential that both susceptible and stable areas be identified to mitigate property damage, environmental degradation, and loss of life. Consequently, landslide susceptibility assessments, i.e., assessments of the spatial probability of a landslide occurring, are a huge step forward in the comprehensive hazard management of landslides [6,7]. The landslide susceptibility map (LSM) produced by a landslide susceptibility assessment can be a useful tool for authorities with decision-making capabilities.Many methods and techniques have been proposed to evaluate landslide susceptibility. In the past few decades, statistical approaches have become popular in the use of remote sensing (RS) with a geographic information system (GIS). There are many statistical approaches used in landslide susceptibility assessment, including a frequency ratio (FR) [8,9], certainty factor (CF) [10], statistical index (SI) [11,12], as well as weight of evidence (WoE) [7,13,14] and logistic regression (LR) [15,16] approaches.Recently, machine learning techniques have become popular in various fields. Machine learning, a branch of artificial intelligence, uses computer algorithms to analyze and predict information based on learning from training data [17,18]. Due to its robustness and high generalization capability, the use of machine learning has increased in landslide susceptibility analysis. Among the machine learning methods, artificial neural network [19,20], fuzzy logic [21,22], neuro-fuzzy [23], support vector machine [24,25], random forest [26,27], and naïve Bayes tree [17,28] methods have been popularly applied.More recently, ensemble machine l...

Landslide Susceptibility Prediction Using Particle-Swarm-Optimized Multilayer Perceptron: Comparisons with Multilayer-Perceptron-Only, BP Neural Network, and Information Value Models

“…Along with the development of information technologies, remote sensing and the geographic information system (GIS) have gradually become data sources and spatial analysis platforms for LSP [6,7]. Based on remote sensing and GIS, many mathematical models have been proposed to calculate landslide susceptibility indices (LSI), such as the analytic hierarchy process [8][9][10], weight evidence method [11], information value (IV) theory [5,12], frequency ratio (FR) method [13,14], logistic regression model [7,15,16], logistic tree model [17], random tree [18,19], boosted tree [20], multi-criteria evaluation model [21], artificial neural networks (ANNs) [22][23][24], support vector machine (SVM) [25][26][27], and neuro-fuzzy method [28]. Although many models have been proposed for LSP, there is no model that is universally accepted and there is much room for improvement for these models.…”

Performance Evaluation of the GIS-Based Data-Mining Techniques Decision Tree, Random Forest, and Rotation Forest for Landslide Susceptibility Modeling

“…In addition, granite gneiss with relatively poor compositional differentiation is excavated en masse, and part of an embedded dike is present. The gneiss outcrop is poor due to severe weathering throughout, and its foliation structure is irregular due to multiple flexures [31,32].…”

“…In addition, granite gneiss with relatively poor compositional differentiation is excavated en masse, and part of an embedded dike is present. The gneiss outcrop is poor due to severe weathering throughout, and its foliation structure is irregular due to multiple flexures [31,32].The soil profile can be divided into three main layers: the colluvium layer, transition zone layer, and clay layer. The colluvium layer is generally composed of loose material, i.e., gravel and silty sand, which extends to a maximum depth of 3.0 m from the surface.…”

GIS Based Novel Hybrid Computational Intelligence Models for Mapping Landslide Susceptibility: A Case Study at Da Lat City, Vietnam

“…In the study, area land use was classified into 3 classes: Agricultural land, non-agricultural land, and barren land (Figure 3d). Landslides are predominant in areas of barren land, on the contrary forest land is less vulnerable to landslides [84]. An NDVI map of land use patterns was created with 5 classes: −0.264-0.1, 0.1-0.2, 0.2-0.3, 0.3-0.6, 0.6-0.645 (Figure 3e).…”