To study the interaction of wind farms with mesoscale effects and with neighboring wind farms we develop and employ high-fidelity large-eddy simulations (LES). These simulations resolve the equations of motion for scales larger than the grid size, while smaller scale motions are modelled with sub-grid scale (SGS) models. Consequently, the accuracy of LES is highly dependent on the SGS model used to parameterize these processes. In Part I we determine the most suitable SGS for large scale simulations. We show that LES using either the anisotropic minimum dissipation (AMD) or the computational more expensive Lagrangian-averaged scale-dependent (LASD) SGS model agrees better with measurements and theoretical predictions than LES using the Smagorinsky model. Furthermore, the effect of selected mesoscale processes is modeled and investigated in microscale LES of wind farms. Here, microscale refers to domain sizes smaller than a few hundred square kilometers. In contrast, mesoscale processes, such as different weather phenomena, cannot be explicitly simulated in microscale domains and must be modeled. Here, we introduce a method to include dynamic wind direction changes, originating from mesoscale atmospheric flow phenomena, in microscale LES. We show that these dynamic wind direction changes can positively and negatively affect the power production of wind farms. Additionally, we include negative geostrophic shear in the LES and show that this phenomenon creates an upward flux above the low-level jet which limits the energy entrainment into the wind farm. In Part II, wind farm wakes and their impact on downstream positioned wind farms are analyzed using LES. We show that the performance of the leading row and the wake recovery of the downstream farm are highly impacted by the wake of the upstream farm. The results are used for an evaluation of the wind farm wake recovery predicted by engineering models. We find that all engineering models under consideration overestimate the wind farm wake recovery compared to LES observations. Therefore, we conclude that these engineering models must be updated to include the interaction between wind farms.