In recent years, the frequency of earthquakes and heavy rainfall has increased, leading to unprecedented damages caused by landcreep, a type of slow-moving landslide. Due to the slow movement of soil layers without rapid topographic changes, landcreep is difficult to detect in advance and often results in significant damage. However, research on landcreep occurrence remained limited in the Republic of Korea. This study assessed the landcreep susceptibility in mountainous forests using the Weight of Evidence (WOE) method, a statistical approach commonly used in natural hazard analysis. With independence tests and the explanatory power of landcreep occurrence factors, five landcreep susceptibility models were developed by combining different factors. The developed models were evaluated by comparing the frequencies of landcreep occurrences according to the landslide susceptibility index (LSI) and the areas under the receiver operating characteristic curves (AUCs). Among the models, Model 3, which incorporated elevation, slope, forest type, bedrock type, distances to faults, roads, and streams, and plan curvature, exhibited the best performance with an AUC of 63.2%. Additionally, the validation efforts using independent datasets resulted in a final AUC of 59.4%. Although the predictive accuracy was relatively moderate, the WOE-based model remains a useful tool for identifying vulnerable areas and supporting disaster management for landcreep in the country.
저자
Chulwon Lee [ Department of Agriculture, Forestry and Bioresources, Seoul National University ]
Corresponding Author
Sangjun Im [ Research Institute of Agriculture and Life Sciences, Seoul National University ]
Song Eu [ Landslide Division, National Institute of Forest Science, 56, Hoegi-ro, Seoul, 02455, South Korea ]
