GIS 기반 통계적 추정 모델을 이용한 산사태 취약성 예측

Abstract

A Study on Prediction of the Landslide Susceptibility by Using Statistically Estimative Models based on GIS

Se Jun Kim

Department of Civil Engineering, The Graduate School, Pukyong National University

Abstract Many regions around the world have experienced more frequent landslides as well as larger typhoons and more severe rainstorms as a result of global warming and abnormally high temperatures. Landslides are thus of increasing concern, and many governments and research institutions have been seeking ways to evaluate and map landslide susceptibility. Geographic information system and remote sensing technologies are used widely to produce landslide susceptibility index maps. However, many researchers nowadays focus on model selection and compare the effects of different models. The quality of model input data, both landslide inventories and triggering factors, continue to be of concern, less than method and technology. Therefore, the present study examined the accuracies of landslide susceptibility index maps produced using various types and combinations of landslide-related factors. For this purpose, Woomyeon Mountain, where severe landslide damage occurred in 2011, was selected as the research area. A database of landslide-related factors was created. The data included landslide locations obtained from aerial photographs, topographic factors derived from airborne LiDAR observations and digital topographic maps, and various soil, forest, and land cover characteristics. Landslide susceptibility index maps were created by calculation of landslide susceptibility index using logistic regression and frequency ratio. The accuracy of the results was calculated using the relative operating characteristic curve. Landslide susceptibility index maps produced using each method had prediction accuracy more than 70%. The landslide susceptibility map produced using frequency ratio had approximately 1~2% higher prediction accuracy than the map produced from logistic regression. In addition to, The prediction accuracy of the landslide susceptibility map produced using airborne LiDAR was improved approximately 2~3% than the map produced from digital topographic maps. As for the correlation between digital elevation map using airborne LiDAR data and the map using digital topographic map, the two data almost matched one another. Landslide susceptibility index maps indicated the existence of a strong correlation. This study shows that airborne LiDAR can aid in compiling correct and systematic databases of landslide-related factors. In addition, the use of airborne LiDAR data results in improved LSI maps, meaning that airborne LiDAR makes it possible to predict the occurrence of landslides more effectively. Precise LSI mapping can aid in disaster planning to minimize casualties and economic losses. In a future study, the enhanced LSI map accuracy shown in this study will be further investigated by using a spatial database including precipitation intensity factors that may affect landslide occurrences and especially debris flows.