로지스틱 회귀분석을 통한 땅밀림지 위험 평가

초록

This study identified landcreep (i.e., a slow-moving landslide) susceptibility in Gyeongsangnam-do and Gyeongsangbuk-do using logistic regression analysis based on elevation, slope, geology, vegetation type, topographic wetness index, slope curvature, distance from waterways, and distance from roads. Spatial analysis was conducted using a 1.5 ha circular buffer for 92 landcreep sites and 92 nonlandcreep sites in the study area. The logistic regression analysis revealed that vegetation type, profile curvature, and distance from roads had statistically significant effects on landcreep occurrence. Model validation yielded a Nagelkerke value of 0.375 and a Hosmer–Lemeshow significance probability of 0.545, suggesting that the model outputs are statistically reliable. Landcreep susceptibility was classified using variables that are statistically significant according to the logistic regression analysis. Grades 1 and 2, representing high-risk areas, accounted for 53.3% of the study area, indicating a moderate prediction rate.

키워드