Application of artificial neural networks and geographic information system to provide hazard susceptibility maps for rockfall failures

• Published in: Environmental earth sciences Vol. 81; no. 19
• Main Authors: Nanehkaran, Y. A., Licai, Zhu, Chen, Junde, Azarafza, Mohammad, Yimin, Mao
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.10.2022; Springer Nature B.V
• Subjects: Accuracy; Algorithms; Analysis; Artificial neural networks; Biogeosciences; Classifiers; Datasets; Decision trees; Distance; Earth and Environmental Science; Earth Sciences; Environmental Science and Engineering; Failure; Failures; Geochemistry; Geographic information systems; Geographical information systems; Geology; Hydrology/Water Resources; Information systems; Landslides; Landslides & mudslides; Learning algorithms; Lithology; Machine learning; Model accuracy; Modelling; Neural networks; Original Article; Prediction models; Remote sensing; Risk; Rock falls; Rockfall; Slopes; Support vector machines; Susceptibility; Terrestrial Pollution; Weathering; Landslides; Rockfall susceptibility assessment; Multilayer perceptron; Artificial neural network; Alborz province
• ISSN: 1866-6280; 1866-6299
• DOI: 10.1007/s12665-022-10603-6

The presented article attempted to analysis rockfalls susceptibility mapping which is considered as one the most important type of the land-slides with high frequent occurrence. The machine learning based multiple-layer perceptron (MLP) model was used to provide the pridicive model and hazard risk maps for studied area. This artificial neural network (ANN) was received significant success in susceptibility assessment for landslides and rockfall. The predictive model was conducted on a dataset containing 15 main triggering factors which concluded 10 regular (e.g., precipitation, slope aspect, slope angle, weathering, lithology, distance to river, main faults’ unsafe radius, landslide-prone areas, distance to roads, and distance to the cities), 5 specific (e.g., RQD, Q slope , GSI, SMR, and RHRS), and 57 historical landslides which represent different aspects of the rockfall failures at Alborz province in Iran. To susceptibility assessment, the dataset that used in ANN-based analysis was randomly divided into training (70%) and testing (30%) sets and utilized to predict the risk-able area.. The predictive model was comparatively justified by using benchmark learning classifiers concluded support-vector machine (SVM), decision tree (DT), and random forest(RF) algorithms. The confusion matrix and receiver operating characteristic curves (ROC) were used for performance analysis and obtaining models’ accuracy. According to the susceptibility results, the north part of the studied region has high risks for rockfall failures. As machine learning-based performance analysis, the MLP-based model by 0.82 accuracy/0.85 precision reached the highest rank more than other classifiers such asSVM (accuracy = 0.78/precision = 0.78), DT (accuracy = 0.70/precision = 0.73), and RF (accuracy = 0.68/precision = 0.70). Also, overall accuracy obtained from ROC, the main model MLP (AUC = 0.811) is higher than SVM (AUC = 0.780), DT (AUC = 0.740), and RF (AUC = 0.500). MLP model provide more conservative results than other classifier and RF is estimated the risk rate less than others.