The optimal rainfall thresholds and probabilistic rainfall conditions for a landslide early warning system for Chuncheon, Republic of Korea

Abstract The purpose of this study is to establish the criteria for a landslide early warning system (LEWS). We accomplished by deriving optimal thresholds cumulative event rainfall–duration (ED) and identifying characteristics rainfall variables associated with high probability occurrence via Bayesian model. have established these using data Chuncheon, Republic Korea. Heavy leading cause landslides in Chuncheon; thus, it crucial determine conditions that trigger landslides. Hourly spanning 1999 2017 from seven gauging stations were utilized ED used three different calibration periods events split 12, 24, 48, 96 non-rainfall hours calibrate thresholds. Finally, threshold was determined comparing results contingency table skill scores maximize detection (POD) score minimize false (POFD) score. In LEWS, considering first level as “normal,” we developed subsequent step-by-step levels based on model well propose second level, “watch,” when condition above then adopt third “warning,” fourth “severe warning,” considers several factors including vs. non-landslide various such hourly maximum 3-day antecedent conditions. proposed alert predicted total 98.2% occurrences at warning” “warning” result fitness verification. alarm rate 0% severe 47.4% level. forecast are likely occur local region. Additionally, ranges represent set standard fits area’s characteristics.