Soil organic carbon (OC) sequestration (i.e. the capture and long-term storage of atmospheric CO2) is being considered as a possible solution to mitigate climate change, notably through land use change (conversion of cropped land into pasture) and conservation agricultural practices (reduced tillage). Subsoil horizons (from 30 cm to 1 m) contribute to ca. half the total amount of soil OC, and the slow dynamics of deep OC as well as the relationships between the OC depth distribution and changes in land use and tillage practices still need to be modelled. We developed a fully modular, mechanistic OC depth distribution model, named OC-VGEN. This model includes OC dynamics, plant development, transfer of water, gas and heat, mixing by bioturbation and tillage as processes and climate and land use as boundary conditions. OC-VGEN allowed us to test the impact of 1) different numerical representations of root depth distribution, decomposition coefficients and bioturbation; 2) evolution of forcing factors such as land use, agricultural practices and climate on OC depth distribution at the century scale. We used the model to simulate decadal to century time scale experiments in Luvisols with different land uses (pasture and crop) and tillage practices (conventional and reduced) as well as projection scenarios of climate and land use at the horizon of 2100. We showed that, among the different tested formalisms/parametrizations: 1) the sensitivity of the simulated OC depth distribution to the tested numerical representations depended on the considered land use; 2) different numerical representations may accurately fit past soil OC evolution while leading to different OC stock predictions when tested for future forcing conditions (change of land use, tillage practice or climate).