Abstract. We present the process and difficulties of acquiring the proper initial and boundary conditions (IBC) for the state-of-the-art LES based model PALM (Parallelized Large-Eddy Simulation Model). We use the mesoscale model WRF (Weather Research and Forecasting model) as a source of inputs for the PALM preprocessor, and investigate the influence of the mesoscale model on the performance of the PALM model. Sixteen different WRF configurations were used as a proxy for a multi-model ensemble. We developed a technique for selecting the suitable sets of IBC, performed PALM model simulations driven by them, and investigated the consequences of selecting a sub-optimal WRF configuration. The procedure was tested for four episodes during different seasons of the year 2019, evaluating WRF and PALM outputs against the atmospheric radio sounding observations. We show that the PALM model outputs are heavily dependent on the imposed IBC, and have different responses for different times of the day, and different seasons. We demonstrate that the main driver of errors is the mesoscale model, and that the PALM model is capable of attenuating, but not fully correcting them. The PALM model attenuates the impact of errors in IBC in wind speed, while for the air temperature, PALM shows variable behavior with respect to driving conditions. This study stresses the importance of high-quality driving IBC, and the complexity of the process of their construction and selection.