LaRaTh-Br: An Event-Based Probabilistic Framework for Modeling Rainfall Conditions Associated with Landslides

Rainfall-based models are widely used to support territorial landslide early warning systems, in which the occurrence of one or more landslides is assessed over predefined warning areas rather than at individual, known slopes. Their development, however, is strongly influenced by methodological choices related to rainfall-event definition, rainfall–landslide association, predictor construction, and model calibration, especially when operational landslide inventories contain uncertainties in the timing and location of reported events. Here, we present LaRaTh-Br (Landslide Rainfall Thresholds—Brazil), a modular R-based computational framework designed to standardize event-based rainfall–landslide analysis. The framework integrates rainfall-event identification from continuous precipitation records, spatial and temporal association between rainfall events and landslide records, generation of multiple rainfall descriptors, and probabilistic modelling and validation. LaRaTh-Br adopts the rainfall event as the fundamental analytical unit and explicitly represents both events associated and not associated with reported landslides. The framework was demonstrated through a retrospective event-level modelling application using ten years of operational rainfall observations and landslide records from Blumenau, southern Brazil. The application showed that event duration and mean rainfall intensity contained most of the discriminatory information, while antecedent rainfall provided complementary information and short-duration peak intensity produced limited additional gain. More importantly, LaRaTh-Br enables alternative rainfall-event definitions, rainfall descriptors, and modelling configurations to be evaluated within a common and reproducible workflow, providing a flexible basis for the continued development and refinement of probabilistic rainfall–landslide models as operational rainfall and landslide datasets expand in Brazil and elsewhere.