A recent study has shown that Doppler lidar is a state-of-the-art method to obtain spatially and temporally resolved flow fields in forest edge flow regimes. In that study, the general flow features observed by lidar were found to be similar to those detected above a physical tree model in a wind tunnel. But in pivotal details, for example regarding the absolute height and the inner structure of the internal boundary layer (IBL), significant differences were detected. The main objectives of this Large-Eddy Simulation (LES) study are to analyze these differences and to associate them to the meteorological and physical differences between the set-ups of the wind tunnel and the atmospheric measurement. This enables on the one hand a model evaluation for the LES and the physical model respectively, and on the other hand a better understanding of the results from the lidar measurements. Results from an LES with neutral stratification and without Coriolis force show a similar IBL structure as in the wind tunnel and represent well-known characteristics of forest edge flow. A variation of the forest density only marginally affects the IBL structure. The presence of a finite forest clearing as observed at the lidar site increases the turbulence level of the IBL, compared to a set-up with a quasi-infinite clearing like in the wind tunnel. Including Coriolis force further enhances the turbulence levels to values observed by lidar. An increasing thermal instability results in even higher turbulence levels. Hence, differences between wind tunnel and atmospheric measurements are mainly traced back to differences in the flow forcing and in the onflow conditions upstream of the forest edge. Furthermore, a statistical analysis reveals that insufficient averaging of the lidar data also contributes to the observed deviations from the wind tunnel results. Based on this analysis, we suggest that at least two and a half hours of measurements during equivalent atmospheric conditions are necessary to obtain a statistically representative mean IBL structure.