Uncertainty in regional scale assessment of landslide susceptibility using various resolutions
NATURAL HAZARDS(2023)
摘要
Resolution produces uncertainty in spatial analysis. The objective of this paper is to study the effects of resolution on landslide susceptibility mapping. First, a landslide survey map that contains 407 historical landslide location information is compiled. In this work, two sampling strategies were used to randomly regroup landslides into two parts for training and testing: one is 70% for training and 30% for testing, whereas the other is 75% for training and 25% for testing. Second, 11 conditioning factors, namely elevation, rainfall, distance to river, plan curvature, aspect, slope, lithology, profile curvature, distance to road, land use, and normalized difference vegetation index, were prepared in ArcGIS version 10.3 for 20 cell sizes from 30 to 600 m with an interval of 30 m. Third, 80 landslide susceptibility maps were generated by combining 20 cell sizes, two sampling strategies, and two models, namely, support vector machine (SVM) and logistic regression (LR). The resolution caused differences in the prediction rates, that is, 6.6–8.2% for SVM and 5.2–8.7% for LR. The best resolutions for the two aforementioned sampling strategies are 150 and 180 m, respectively. The optimal resolution should be related to the landslide size and close to the average area of the landslide when the landslide inventory map is presented by landslide points. This study provides a reference for the resolution comparison in landslide assessment and enhances a new understanding of the relationship between optimal resolution and landslide size.
更多查看译文
关键词
Landslide susceptibility mapping,Loess area,SVM,LR,DEM resolution
AI 理解论文
溯源树
样例
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要