Rainfall thresholds are the most used tool for predicting rainfall-induced thresholds worldwide. Nevertheless, a limited landslide catalogue hinders the definition of rainfall thresholds. Moreover, rainfall thresholds solely consider the rainfall characteristics and overlook another widely recognized indicator for landslide initiation, antecedent soil moisture condition. Thus, this study aims at defining credible hydro-meteorological thresholds for Lueyang county, Shaanxi Province using a landslide catalogue that exhibits a great imbalance between landslide occurrences and non-occurrences. The cumulated rainfall (E) and rainfall duration (D) thresholds were defined at several exceedance probability levels, considering the rainfall events not associated with landslides. The Bayesian approach was exploited to estimate landslide occurrence probabilities within the conditions of rainfall severity and antecedent soil moisture, where rainfall severity is determined by defined E-D thresholds, and soil moisture information is retrieved from a remote sensing reanalysis dataset (ERA5-Land). The results show that the defined negative E-D thresholds exhibit good reliability and robustness. Furthermore, by considering antecedent soil moisture, the predictive capability of hydro-meteorological thresholds shows a significant improvement.
|