Abstract. For rainfall-induced landslides, their occurrence is attributed to both the antecedent wetness condition and the recent rainfall condition. However, when defining rainfall thresholds for the landslide occurrence, these two types of information have been used incompletely or implicitly, which may affect the threshold's predictive capability. This study aims to investigate how to make a better use of these two types of information in the landslide threshold definition. Here four types of thresholds are proposed, by including different variables that are responsible for landslide occurrences, these thresholds could represent different cases, like whether to include the antecedent wetness information or whether to consider the recent rainfall condition explicitly. The predictive capability of these thresholds is then compared crossly with the help of the receiver operating characteristic (ROC) approach. We carry out this study in a northern Italian region called Emilia-Romagna. Results show that the antecedent wetness condition plays a crucial role in the occurrence of rainfall-induced landsides. It is beneficial for the threshold's predictive capability to explicitly include the antecedent wetness information and the recent rainfall in the definition of landslide thresholds. When including soil moisture information in landslide threshold, the reliability of the soil moisture measurement is a key factor affecting the threshold's prediction performance.