Assessment of Landslide Susceptibility of the Wiśnickie Foothills Mts. (The Flysch Carpathians, Poland) Using Selected Machine Learning Algorithms

Abstract

Landslides are well-known phenomena that cause significant changes to the relief of an area’s terrain, often causing damage to technical infrastructure and loss of life. One of the possible means of reducing the negative impact of landslides on people’s lives or property is to recognize areas that are prone to their occurrence. The most common approach to this problem is preparing landslide susceptibility maps. These can factor in the actual location of landslides or the causal relationship between landslides and selected environmental factors. Creating a classification of landslide-prone areas is a challenging task when landslide density is not high and the area of analysis is large. We prepared shallow 10 m × 10 m resolution landslide susceptibility maps of the Wiśnickie Foothills (Western Carpathians, Poland) using eleven different machine learning algorithms derived from the Python libraries Scikit-learn and Imbalanced-Learn. The analyzed area is characterized by a mean density of 3.4 surficial landslides (composed of soils and rocks) per km2. We also compared different approaches to imbalanced sets of data: Logistic Regression, Naive Bayes, Random Forest, AdaBoost, Bagging, ExtraTrees (Extremely Randomized Trees), Easy Ensemble, Balanced Bagging, Balanced Random Forest, RUSBoost and a hybrid model combining Random Under Sampler and Multi-layer Perceptron algorithms. The environmental factors (slope inclination and aspect, distance from rivers, lithology, soil type and permeability, groundwater table depth, profile and plan curvature, mean annual rainfall) were categorized and divided into training (70%) and testing (30%) sets. Accuracy, recall, G-mean and area under receiver operating curve (AUC) were used to validate the quality of the models. The results confirmed that algorithms based on decision tree classifiers are suitable for preparing landslide susceptibility maps. We also found that methods that generate random undersampling subsets (Easy Ensemble, Balanced Bagging, RUSBoost) and ensemble methods (Bagging, AdaBoost, Extra-Trees) both yield very similar test results to those that use full sets of data for training. Relatively high-quality results can also be obtained by integrating the Random Under Sampler algorithm with the Multi-layer Perceptron algorithm.

Landslide type inference based on statistical analysis of a high-resolution digital elevation model in Gorce National Park, Poland

“…The proposed method can be easily scaled and automated, which is particularly important in the case of the rapidly growing number of high-resolution digital terrain models. With the increasing use of machine learning in landslide analysis, quantitative information about landslide geometry is important as it makes it easier to define trigger mechanisms 55 , 56 . Statistically significant differences in the structural context of landslides found for the study area create opportunities for the practical use of this information in other analyses.…”

Improving landslide susceptibility prediction through ensemble Recursive Feature Elimination and meta-learning framework

“…The division was done randomly. Empirical research consistently demonstrates that allocating 20-30% of the data for testing and the remaining 70-80% for training yields optimal outcomes (Akıncı, 2022;Ye et al, 2022;Zydroń et al, 2022;Chen & Fan, 2023;Guo et al, 2023;). In order to address the problem of underfitting and overfitting caused by the dataset size, this work conducted a series of experiments with different proportions of training and testing data (70:30, 75:25, and 80:20).…”

“…These methods build models based on the assumption that the conditions leading to landslides have causal relationships with historical events. Diverse algorithms have effectively been used in the investigation of creating landslide susceptibility maps: Adaptive Boosting (AdaBoost) (Jennifer, 2022;Zydroń et al, 2022), Artificial Neural Network (ANN) (Achu et al, 2023), Convolutional Neural Network (CNN) (Chen & Fan, 2023;Habumugisha et al, 2022), Deep Neural Network (DNN) (Achu et al, 2022;Habumugisha et al, 2022), MultiLayer Perceptron Neural Network (MLP) (Liu et al, 2021;Guo et al, 2023), However, the utilization of a meta classifier, instead of relying solely on individual models, always gives a strategic advantage in improving the predictive performance and robustness of complex machine learning tasks.…”

“…In the present study, six advanced machine learning (ML) models including Logistic Regression (LR) (Chen & Fan, 2023;Khaliq et al, 2023), Support Vector Machine (SVM) (Yang et al, 2023), Random Forest (RF) (Achu et al, 2023;Shoaib et al, 2023), Extremely Randomized Trees (ExtraTrees) (Zydroń et al, 2022;Zhou et al, 2023), Gradient Boosting (GB) (Akıncı, 2022;Shahzad et al, 2022), Extreme Gradient Boosting (XGBoost) (Ye et al, 2022;Cao et al, 2023;Guo et al, 2023;Janizadeh et al, 2023) and a Meta Classifier (MC), which integrates the prediction of six distinct models, were adopted to construct landslides 4 susceptibility maps using advanced Remote Sensing (RS) and Geographic Information System (GIS) techniques. The implementation of an ensemble feature selection process in our study reflects a pioneering approach to landslide susceptibility mapping.…”