Over the past 9,150 years, at least 9 flank collapses have been identified in the history of La Soufrière of Guadeloupe volcano. On account of the volcano’s current unrest, the possibility of such a flank collapse should not be dismissed in assessing hazards for future eruptive magmatic as well as non-magmatic scenarios. We combine morphological and geophysical data to identify seven unstable structures (volumes ranging from 1 × 106 m3 to 100 × 106 m3), including one that has a volume compatible with the last recorded flank collapse in 1530 CE. We model their dynamics and emplacement with the SHALTOP numerical model and a simple Coulomb friction law. The best-fit friction coefficient to reproduce the 1530 CE event is tan(7°) = 0.13, suggesting the transformation of the debris avalanche into a debris flow, which is confirmed by the texture of mapped deposits. Various friction angles are tested to investigate less water-rich and less mobile avalanches. The most densely populated areas of Saint-Claude and Basse-Terre, and an area of Gourbeyre south of the Palmiste ridge, are primarily exposed in the case of the more voluminous and mobile flank collapse scenarios considered. However, topography has a prominent role in controlling flow dynamics, with barrier effects and multiple channels. Classical mobility indicators, such as the Heim’s ratio, are thus not adequate for a comprehensive hazard analysis.

The propagation of rapid gravity-driven flows (Iverson & Denlinger, 2001) occurring in mountainous or volcanic areas is a complex and hazardous phenomenon. A wide variety of events are associated with these flows, such as rock avalanches, debris avalanches and debris, mud or hyper-concentrated flows (Hungr et al., 2014). The understanding and estimation of their propagation processes is important for sediment fluxes quantification, for the study of landscapes dynamics. Besides, gravity-driven flows can have a significant economic impact and endanger local populations (Hungr et al., 2005;Petley, 2012;Froude & Petley, 2018). In order to mitigate these risks, it is of prior importance to estimate the runout, dynamic impact and travel time of potential gravitational flows.This can be done empirically, but physically based modeling is needed to investigate more precisely the dynamics of the flow, in particular due to the first-order role of local topography. Over the past decades, thin-layer models (also called shallow-water models) have been increasingly used by practitioners. Their main assumption is that the flow extent is much larger than its thickness, so that the kinematic unknowns are reduced to two variables: the flow thickness and its depth-averaged velocity. The dimension of the problem is thus lower, allowing for relatively fast numerical computations. The first and simplest form of thin-layer equations was given by Barré de Saint-Venant (1871) for almost flat topographies. The 1D formulation (i.e., for topographies given by a 1D graph Z = Z(X)) for any bed inclination and small curvatures was derived by Savage and Hutter (1991). This model has since been extended to real 2D topographies (i.e., given by a 2D graph Z = Z (X, Y)). Some of the software products based on thin-layer equations are currently used for hazard assessment to derive, for instance, maps of maximum flow height and velocity. Examples include RAMMS (