Karthala, on Grande Comore Island, is a basaltic shield volcano characterised by sporadic ash-forming phreatic eruptions. In 2005, two of such mildly explosive episodes emplaced large quantities of tephra on the summit area, initiating the repetitive occurrence of small-scale secondary lahars up to 2012. In this study, we aim to characterise and model these rain-triggered flows which heavily impacted the settlements situated at the foot of Karthala, damaging infrastructures and affecting thousands of inhabitants. We mapped the lahar deposits and surveyed the morphology of eight flow-guiding ravines in the field during the summer of 2015. We compared for the first time numerical lahar simulations from the widely used LAHARZ with results from Q-LavHA, a recently developed probabilistic model initially designed for lava flow emplacement modelling. The depositional extent mapped in the field served as input for a fitness index which assessed the accuracy of the resulting simulations. Additionally, we updated the virtual topography (Digital Elevation Model) that is used as model input, by considering the measurements of the ravine geometry acquired in the field and assessed its impact on the accuracy of the simulated lahars. Our study allows to define the flows as small-scale, rain triggered and low sediment concentration lahars, i.e. with a total volume of 2-5.105 m3 and a peak discharge of ca. 10 m3/s. The comparison of the simulation outputs shows that Q-LavHA is a suitable model for such small-scale lahar simulations, with accuracy values similar or higher to the ones obtained by LAHARZ. Its strengths lie principally in its accurate simulation of the lateral spread of the flow downslope from concave break-in-slopes and its ability to model flow bifurcation. Our results also indicate an accuracy increase, with more accurate flow trajectories, of the modelled outputs of LAHARZ and Q-LavHA when detailed channel topography is used to update the virtual topography on a small spatial scale. The numerical simulation results highlight their added value as part of an integrated volcanic hazard assessment. Based on this case study, we recommend the use of Q-LavHA and updated virtual topography as the base for the production of lahar inundation hazard maps for highly populated volcanic islands such as Grande Comore.