期刊
BULLETIN OF ENGINEERING GEOLOGY AND THE ENVIRONMENT
卷 74, 期 1, 页码 27-45出版社
SPRINGER HEIDELBERG
DOI: 10.1007/s10064-014-0607-7
关键词
Landslide susceptibility; GIS; Landslide susceptibility index; Artificial neural networks; Peloponnesus; Greece
The main scope of this study is to compare the performance of a conventional statistical method like the landslide susceptibility index (LSI) and a soft computing method like artificial neural networks (ANNs). These models were applied in order to realistically map landslide susceptibility (LS) in the Krathis and Krios drainage basins in northern Peloponnesus. The relationship between landslides and various conditioning factors contributing to their occurrence was investigated through geographic information system-based analysis. A landslide inventory was realised using aerial-photos, satellite images and field surveys. Eight conditioning factors, including land cover, geology, elevation, slope, aspect, distance to road network, distance to drainage network, distance to structural elements, were considered. Subsequently, LS maps were produced using LSI and ANNs, and they were then compared and validated accordingly. Model performance was checked by an independent validation set of landslide events. For the validation process, the receiver operating curve was drawn and the area-under-the-curve (AUC) values were calculated. The calculated AUC values were 0.852 for the LSI model, and 0.842 for the ANNs; thus, both methods seem to lead to quite similar results. Based on these results, with an average percentage of correctly predicting landslides of about 84 %, model validation confirms that extrapolation results are very good, and that both models can be used to mitigate hazards related to landslides, and to aid in generalised land-use planning assessment purposes.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据