Phase-resolving Direct Numerical Simulations of bed-load transport in a horizontal turbulent open-channel flow at small relative submergence are presented. The thesis provides a detailed study addressing the impact of the choice of collision model on the scenario of bed-load transport and presents statistical tools to identify and describe the key-mechanisms governing the fluid-particle interaction. The Double-Averaging Methodology is applied for the first time to the situation of mobile rough beds. This methodology provides a framework to convolute the data in such a way that the most prominent flow features are well described by a handy set of double-averaged (in time and space) quantities. The thesis further provides a systematic study elucidating in detail the impact of the key-parameters mobility and sediment supply on the pattern formation of large-scale particle clusters. This is done using a very large computational domain to allow bed-forms to evolve with minimal spatial constraints. It is found that a low transport rate is linked to streamwise oriented ridges, while a large sediment supply results in large-scale clusters that propagate in streamwise direction. A detailed description of fluid quantities links the developed particle patterns to the enhancement of turbulence. The large domain allows for a large number of independent erosion events, such that conditional averaging provides a very clear description of the processes involved for incipient particle motion. Furthermore, the detection of moving particle clusters as well as the investigation of their surrounding flow field is performed by an analysis using a moving frame coordinate system.