All surfaces, whether they are natural or man-made, exhibit a certain amount of roughness on a range of length scales. This surface roughness evidently plays a major role in tribological processes, like friction and wear between two surfaces sliding against each other. While those processes have important implications in engineering or in geology, it is currently not possible to accurately predict them. Only empirical models exist, which must be fitted against experimental results and are only valid on a limited range. The limited success of empirical modeling is a motivation to move toward a more fundamental bottom-up approach. Using molecular dynamics (MD) simulations, it is possible to simulate the onset of wear at the scale of atomic asperities [1], allowing the establishment and validation of a wear law for single asperities. With the same kind of simulations, more complex sliding rough surfaces can be simulated in 2D to study the formation and evolution of third body particles and worn surfaces [2]. Going toward more realistic three-dimensional setups would be however very computationally expensive using the same MD framework, due to the large number of simulated atoms required. Instead, the discrete element method (DEM) can be used to greatly mitigate this computational cost. Using DEM particles with contact forces and cohesive forces, it is possible to simulate a deformable solid while having particle diameters and system sizes an order of magnitude higher than with MD. The pairwise forces can be tuned to obtain a solid with reasonably approximated elastic and fracture properties. The simulations of single asperity wear performed with MD can be successfully reproduced with DEM within a range particle sizes, validating the coarse graining procedure. More complex simulations allow us to study the formation of third body particles and the evolution of worn surfaces in an adhesive wear context.