1 INTRODUCTION Erosion, whether internal or surface, is one of the main mechanisms leading to the breaching of em- bankment dikes and dams. A reliable determination of the erodibility of soils is then necessary. There- fore some erosion tests devices were introduced, such as the Hole Erosion Test (Wan & Fell 2004) and the Jet Erosion Test (Hanson & Cook 2004), two of the most commonly used apparatus. They are designed to grade the sensitivity of soils to erosion in the laboratory or in-situ by performing standard- ized tests. allow to answer the following three ques- tions about the different materials tested: When is the erosion triggered? What is the speed of degrada- tion by erosion? When does the erosion stop? The determination of erosion parameters: the erosion threshold or critical shear stress ( t c ) and the erosion coefficient ( d k ), permit answering these questions. Nevertheless, the values of the erosion parameters obtained following these two tests present major dif- ferences, as highlighted by Regazzoni et al. (2008) and Wahl et al. (2008). To validate the interpretation model of the Jet Erosion Test (JET) and lead to a better understanding of the erosion processes, the development of a numerical model able to predict the erosion of a cohesive soil by turbulent flow is necessary. At the state of art, none of the validated numerical models of erosion (Vardoulakis et al. 1996 and Ouriemi et al. 2009 & Lachouette et al. 2008) permits to model the erosion of a cohesive soil for turbulent flow cases, except from the numerical modelling method developed by Mercier et al. (2014). This numerical model is able to predict the erosion of a cohesive soil by turbulent flow, in case of impinging jets. This study presents the extension of the Mercier et al. (2014) numerical model, to tan- gential flows. Section 2 details the numerical model. In section 3, a theoretical validation of the modelling method is exposed. Then, in section 4, the erosion process dur- ing Hole Erosion Test (HET) is modelled numerical- ly. Comparison between experimental and numerical results is performed. A discussion of the results is finally proposed in section 5. 2 MODELLING METHOD 2.1 Challenges The first difficulty that becomes apparent in this context is taking into account the two-phase nature On the numerical modelling of the Hole Erosion Test F. Mercier, S. Bonelli & P. Philippe IRSTEA, Aix-en-Provence, France F. Golay Université de Toulon, La Garde, France F. Anselmet IRPHE, Marseille, France P. Pinettes geophyConsult, Savoie Technolac, France J.J. Fry EDF-CIH, Le Bourget du Lac, France ABSTRACT: This study focuses on 2D Computational Fluid Dynamics (CFD) numerical modelling of the erosion of a cohesive soil by turbulent flows. A numerical model of erosion has been developed, based on adaptive remeshing of the water/soil interface. The flow is modelled according to the Reynolds Averaged Na- vier Stokes (RANS) method, which closure problem is solved with the k-ε turbulence model. A threshold shear stress erosion law is implemented. The model is first validated with a theoretical benchmark, namely the erosion of a channel subject to a laminar flow in a 2D configuration. The numerical results are in satisfac- tory agreement with the theoretical prediction, within less than 2%. Then three Hole Erosion Tests (HETs) performed on different soils are modelled numerically. The soil’s erosion parameters injected in our model are those deduced from the HET interpretation model of Bonelli et al. (2006). The numerical model is now developed with a 2D axisymmetrical configuration in turbulent conditions. The results obtained for three dif- ferent HET tests are in good agreement compared to the experimental data. Our model turns out to be able to reproduce accurately the erosion of a cohesive soil subject to a concentrated leak and gives additionally ac- cess to a detailed description of all the averaged hydrodynamic flow quantities. Such a detailed description is essential for a better understanding of the erosion processes.