Assessing two methods of defining rainfall intensity and duration thresholds for shallow landslides in data-scarce catchments of the Colombian Andean Mountains

Rainfall thresholds are intensity-duration relations that are supposedly able to distinguish precipitation events that may or may not trigger landslides. The most common method for defining rainfall thresholds relies on observed landslides and the corresponding values of rainfall intensity and duration that caused each failure. Alternative methods to define rainfall thresholds, using physically-based models, recently gained importance, as they may provide complementary information to other methods. Still, their applicability in most of the world's regions, including the mountainous basins of the Colombian Andes, has not been demonstrated or validated. In this study, we evaluated the applicability of the physically-based model TRIGRS to define rainfall intensity and duration thresholds in individual basins from the Colombian Andes. We obtained rainfall thresholds using two different methods and compared them with landslide-triggering rainfall events in two distinct basins, namely La Arenosa and La Liboriana. Furthermore, we used a (presumably incomplete) landslide database from Medellin to rebuild the rainfall events associated with individual landslides and compared them with the physically-based thresholds. The rainfall thresholds calculated in the three study areas and the applicability of the methods in data-scarce environments were assessed. Results showed that both methods for defining rainfall intensity and duration thresholds have merits and represent potential tools for improving or complementing landslide early warning systems, especially in data-scarce regions.