To quantify and evaluate the effect of root reinforcement on slope stability, it is of primary importance to understand the mechanical interaction between roots and soil. At the slope scale, root reinforcement is a key input when performing stability analysis using either Limit Equilibrium Method or Finite Element Method based models. In nearly all the previous approaches, soil mechanical reinforcement by the roots was modelled as a single additional cohesion to the soil effective cohesion. This modelling approach allowed simplified integration of roots impact on slope stability and made stability models easy to compute. Nevertheless, the hypothesis of considering the effect of roots as an additional cohesion term has been increasingly challenged by studies conducting soil-roots shear tests at a local scale, i.e. the scale of the root system or root bundle embedded in soil. Although the experimental approaches provide an effective way to study root-soil interaction, they are usually time-consuming and laborious, especially for generating replicates, considering complex, multiple, and correlated root traits, and controlling environmental factors. Compared to experimental tests, numerical simulations at the plant scale constitute a promising alternative to study roots-soil interaction. Despite difficulties in model validation using field data, simulations based on the Finite Element Method can be considered as reliable numerical approaches generating very comparable results with experimental tests. Recently, a rooted soil modelling approach based on the Discrete Element Method was also developed and showed powerful potentials to explore complex root-soil interactions. In this research work, we simulated 3D direct shear tests using the standard implicit Finite Element Method (FEM) and he Discrete Element Method (DEM), aiming at (i) comparing the two numerical approaches and (ii) evaluating classical soil reinforcement models. For that purpose, in homogeneous soil with low cohesion, 36 straight, non-branched and thin root models were implanted in three parallel lines. Root traits, including orientation with regards to the shear strain direction (45°, 90° and -45°), longitudinal modulus of elasticity (10 MPa and 100 MPa), and bending and compressive root behaviours (beam, truss and cable) were investigated. The results from this analysis clearly showed that, compared to the FEM, the DEM achieved consistent results, avoided convergence problems, but required longer computation time and used parameters potentially difficult to identify. In addition, the DEM presents the advantage of a more detailed modelling of the root soil local interaction and, thus of root slippage into the soil. Both advantages and drawbacks of each approach tend to show the necessity of using both of them as complementary tools in future studies. The results also showed that root reinforcement varied as a function of soil strain and was closely related to root geometry, location in the sample, and mechanical traits. In addition, existing root reinforcement models tended to provide higher root reinforcement estimates than those achieved by numerical direct shear tests. As suggested by other recent research studies, the results highlight the necessity to take into account the effect of soil strain, root orientation, position, root type and confining normal pressure in the existing root reinforcement models in order to enhance both their accuracy of prediction and their closeness to the modelled or observed processes.