Abstract/Summary

The availability of suitable and reliable reference data together with the application of modelspecific comparison methods are the essential ingredients to establish confidence in the capabilities of a numerical model and to truly assess its strengths and limitations. This thesis is motivated by the striking lack of proportion between the increasing use of large-eddy simulation (LES) as a standard modeling technique in micro-meteorological research as opposed to the level of scrutiny that is commonly applied to the quality of the generated numerical predictions. With this study, I suggest and apply a novel validation strategy for LES consisting of a multilevel hierarchy of comparative analysis methods. Unlike standard LES validation procedures that are based on the comparison of low-order statistical moments, the new approach advocated here specifically aims at the time-dependent nature of the problem. The sequence in which statistical quantities are compared mirrors the increase of information provided by the analysis methods. The target area is turbulent flow in the near-surface atmospheric boundary layer. The test scenario for the validation approach is urban flow in the city of Hamburg, Germany. Qualified reference data are generated in the boundary-layer wind tunnel facility at the University of Hamburg through high-resolution flow measurements in a scale-reduced model. Fine-meshed numerical simulations are conducted at the U.S. Naval Research Laboratory in Washington, D.C., with implicit LES. On the basis of an initial exploratory data analysis of mean flow and turbulence statistics, a high level of agreement between simulation and experiment is apparent. Inspecting frequency distributions of the underlying instantaneous data, however, proves to be necessary for a more rigorous assessment of the overall prediction quality. From histograms, local accuracy limitations caused by under-resolution as well as particular strengths of the model to capture complex urban flow features are readily determined. Further crucial information about the physical validity of the LES need to be obtained from eddy statistics. Comparisons of temporal autocorrelations, integral time scales, and auto-spectral energy densities show that the simulation reliably reproduces statistical characteristics of the energy and flux-carrying roughness sublayer structures. At higher elevations, however, inflow generation artifacts are reflected in dubiously short fluctuation time scales and energy peaks that are dislocated toward high frequencies. With the comparison of scale-dependent flow statistics, to which the preceding diagnostics have been blind, the emphasis eventually shifts to structure identification. The quadrant analysis of the vertical turbulent momentum flux discloses strong similarities between ejection-sweep patterns and the occurrence of rare, but extreme, flux events in roof-level vicinity and above the canopy layer. Further scale-wise comparisons of wavelet-coecient frequency distributions and associated high-order statistics reveal consistent location-dependent intermittency patterns induced by eddies in the energy-production range. Compared with usual methods that rely on single figures of merit, the detailed, multi-level validation strategy presented in this thesis allows to draw more wide-ranging and tenable conclusions about the quality of the simulation and to specify the model’s fitness for purpose in greater detail. The proposed validation concept has the potential to be used as a starting point for communitywide activities aiming at the formulation and harmonization of best-practice standards for the quality assurance of micro-meteorological eddy-resolving simulations.