The energy transfer from a laser beam source to material surfaces with arbitrary geometrical features and variable surface roughness is the crucial step in many high-end engineering applications. We propose two models capable of predicting this energy transfer, applicable in different scenarios. The first is a high-fidelity numerical framework for the simulation of laser beam interaction with rough surfaces, which includes meshed geometry of arbitrary shape and material Lagrangian particles. The method discretizes the laser source as a collection of photon-type immaterial Lagrangian particles (Discrete Element Method) and is able to capture the effects of multiple reflections, angle-dependent reflectivity, and polarization change. Simulations were performed on a geometry reconstructed from a rough copper sample to reveal the impact of the polarization effects. This method is generally applicable to any surface where the effects of inelastic light scattering are not expected to play a significant role. The second model is a novel phenomenological correlation specifically designed to predict the effective reflectivity of sparse powder layers, which occur for example when metal vapor is recondensed and redeposited on the substrate during laser welding. The correlation is compared to the predictions obtained from the simulation framework and has been favorably compared to experimental data in a separate publication.