Methods of eddy structure identification are applied to velocity data of atmospheric surface layer flows modeled in a boundary-layer wind tunnel. The objective is to test their potential to serve as mathematical tools for the validation of eddy-resolving numerical models like large-eddy simulation and for the generation of realistic turbulent inflow conditions. The reconstruction of complex atmospheric flows on the basis of two-point space-time statistics is tested with the proper orthogonal decomposition and linear stochastic estimation that are both applied to spatially well-resolved flow data. The continuous wavelet transform is used to derive joint time-frequency information from single-point velocity time series. Whereas the proper orthogonal decomposition and the continuous wavelet transform show particular strengths in the spatiotemporal characterization of turbulent flows, the stochastic estimation is moreover qualified to generate new flow scenarios from a minimum number of instantaneous data. (C) 2011 Elsevier Ltd. All rights reserved.