It is assumed that the description of the exchange processes between heterogeneous natural surfaces and the atmosphere within turbulence closure models is mainly limited by a realistic three-dimensional (3D) representation of the vegetation architecture. Within this contribution we present a method to record the 3D vegetation structure and to use this information to derive model parameters that are suitable for numerical flow models. A mixed conifer forest stand around a clearing was scanned and represented by a dense 3D point cloud applying a terrestrial laser scanner. Thus, the plant area density (PAD) with a resolution of one cubic meter was provided for analysis and for numerical simulations. Multi-level high-frequency wind velocity measurements were recorded simultaneously by 27 ultrasonic anemometers on 4 towers for a period of one year. The relationship between wind speed, Reynolds stress and PAD was investigated and a parametrization of the drag coefficient CD by the PAD is suggested. The derived 3D vegetation model and a simpler model (based on classical forest assessments of the site) were applied in a boundary layer model (BLM) and in large-eddy simulations (LES). The spatial development of the turbulent flow over the clearing is further demonstrated by the results of a wind tunnel experiment. The project showed, that the simulation results were improved significantly by the usage of realistic vegetation models. 3D simulations are necessary to depict the influence of heterogeneous canopies on the turbulent flow. Whereas we found limits for the mapping of the vegetation structure within the wind tunnel, there is a considerable potential for numerical simulations. The field measurements and the LES gave new insight into the turbulent flow in the vicinity and across the clearing. The results show that the zones of intensive turbulence development can not be restricted to the locations found in previous studies with more idealized canopies.