Large-Eddy Simulation (LES) has become a common tool to investigate various topics in micro-meteorology. However, an evaluation of LES models using adequate experimental data is still necessary. Previous comparisons between simulated and experimental data have often been reduced to vertical profiles of mean quantities and were only qualitatively useful. Within this project the question of evaluating the three-dimensional, time-dependent, random velocity and temperature fields from LES using acoustic tomography was examined. The experimental method of acoustic travel time tomography provides two- or three-dimensional arrays of temperature data, averaged over grid cells. Because of this format, tomography data is suitable for a comparison with LES-data.

A first comparison of acoustic tomography measurements and simulated data within a convective boundary layer showed qualitative correlation. However, some quantitative deviation of experimental data with LES-data could be observed. One possible reason for quantitative deviations could be the accuracy of initial and boundary conditions for the LES (i.e. the initial temperature profile, the roughness length, and the geostrophic wind speed). A sensitivity study performed with the LES-model allowed to quantify the influence of uncertainties in the initial model parameters on the simulation results. It shows that these initial and boundary conditions have to be provided with an accuracy that could hardly be reached by experimental methods. One crucial parameter is the surface roughness length, which plays an important role for the computed heat flux in the model and therefore for the simulated convective boundary layer development. For quantitative comparisons its inaccuracy should not exceed +-30%. However, such precise values can hardly be provided by state-of-the-art experiments, especially when the surface is inhomogeneous.