Evaluation of the Importance of Variables When Using a Random Forest Technique to Assess Landslide Damage: Focusing on Chungju Landslides; Random Forest를 활용한 산사태 피해 영향인자 평가: 충주시 산사태를 중심으로

초록

Landslides are natural disasters that causes significant property damage worldwide every year. In Korea, damage due to landslides is increasing owing to the effects of climate change, and it is important to identify the factors that increase the prevalence of landslides in order to reduce the damage they cause. Therefore, this study used a random forest model to analyze the importance of 14 factors in influencing landslide damage in a specific area of Chungju, Chungcheongbuk-do province, Korea. The random forest model performed accurately with an AUC of 0.87 and the most-important factors were ranked in the order of aspect, slope, distance to valley, and elevation, suggesting that topographic factors such as aspect and slope more greatly influence landslide damage than geological or soil factors such as rock type and soil thickness. The results of this study are expected to provide a basis for mapping and predicting landslide damage, and for research focused on reducing landslide damage. © 2024 Korean Society of Engineering Geology. All rights reserved.

키워드