Landslide susceptibility mapping (LSM) is a useful tool to estimate the probability of landslide occurrence, providing a scientific basis for natural hazards prevention, land use planning, and economic development in landslide-prone areas. To date, a large number of machine learning methods have been applied to LSM, and recently the advanced convolutional neural network (CNN) has been gradually adopted to enhance the prediction accuracy of LSM. The objective of this study is to introduce a CNN-based model in LSM and systematically compare its overall performance with the conventional machine learning models of random forest, logistic regression, and support vector machine. Herein, we selected Zhangzha Town in Sichuan Province, China, and Lantau Island in Hong Kong, China, as the study areas. Each landslide inventory and corresponding predisposing factors were stacked to form spatial datasets for LSM. The receiver operating characteristic analysis, area under the curve (AUC), and several statistical metrics, such as accuracy, root mean square error, Kappa coefficient, sensitivity, and specificity, were used to evaluate the performance of the models. Finally, the trained models were calculated, and the landslide susceptibility zones were mapped. Results suggest that both CNN and conventional machine learning-based models have a satisfactory performance. The CNN-based model exhibits an excellent prediction capability and achieves the highest performance but also significantly reduces the salt-of-pepper effect, which indicates its great potential for application to LSM.
There are 51 tributaries in the middle reaches of the Yarlung Zangbo River (YZR), and the confluences of 87% of the tributaries west of Jiacha Gorge are high-angle or perpendicular, reflecting the anomalous development of these tributaries. In this paper, field investigation and digital elevation model (DEM) methods were used to analyse the causes of this anomalous phenomenon, and it was found that there was a watershed in the area of the Jiacha Gorge. The palaeo-YZR west of the Jiacha Gorge flowed westward before the early Pleistocene into the Zada, Zhongba, Jilong and Gamba–Dingri palaeolakes, which featured a large amount of total accommodation space in the western Qinghai–Tibet Plateau; thus, this river was a continental river. With the intensification of the collision between the Indian plate and the Eurasian plate, the Qinghai–Tibet Plateau experienced rapid uplift and formed a landscape with high elevations in the west and lower elevations in the east, promoting the headward erosion of the eastward-flowing river. During the early Pleistocene, the river east of the Jiacha Gorge crossed the watershed and captured the palaeo-YZR, causing a reversal in the flow direction of the palaeo-YZR.
Landslide susceptibility assessment plays a critical role in disaster management and post-disaster planning. Machine learning-based approaches have recently attracted a lot of attention. However, the parameters tuning in this category of methods has not been accurately determined and is even considered as a weak point. The main objective of this study is to develop two machine learning-based landslide susceptibility models that optimized using a metaheuristic optimization algorithm, the grey wolf optimizer (GWO), for assessing the probability of landslide occurrence without artificial tuning. The selected machine learning algorithm were random forests (RF) and support vector machines (SVM). We apply the optimized models to Jiuzhaigou County on the eastern margin of Qinghai-Tibet Plateau. A total of 270 earthquake-triggered landslides were identified by remote sensing interpretation and filed surveys. Sixteen predisposing factors involving geology, human activity, and hydrology were extracted from the available materials. Then thirteen factors suitable for the study area were selected using multicollinearity diagnosis methods. Two meta-optimization models, GWO-RF, GWO-SVM, were con-structed after GWO's automated search for model parameters. Finally, the Receiver Operating Characteristic (ROC) curve and related statistics, including Accuracy, Sensitivity, and Specificity, were chosen to evaluate and compare the performance of the optimized landslide susceptibility models. Both models were constructed with ROCs higher than 0.95 on the training dataset and validation dataset as well as high accuracy. GWO-RF obtained the best both of accuracy and AUC values of 0.9198 and 0.972 on the validation dataset, respectively. Furthermore, we performed a weighting analysis of the factors and speculated on the relationship between the raw data distribution and accuracy. The results of this study show that the construction of the landslide susceptibility model optimized using a metaheuristic optimization algorithm is a feasible approach.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.