This paper presents the strengths and weaknesses of an earlier-proposed framework based on perturbation theory for extracting approximate analytical models of airwake from a database. Accordingly, the framework is tested against a comprehensive database of autospectra from both experimental and CFD-based investigations. The database covers a wide range of flow topology at various locations under different superstructure and wind-over-deck conditions. The autocorrelations for all three velocity components are represented by perturbation series in which the first terms have a form of the von Karman point correlation functions, each velocity component having its own time scale and rms value. These series are then transformed into equivalent perturbation series of autospectra. The perturbation coefficients are determined by satisfying the analytical constraints in the formulation and through a nonlinear curve-fitting on a set of selected data points in the low-frequency bandwidth (0 ≤ ω rad/s ≤ 10). The extracted models facilitate qualitative analysis as interpretive models; generally they are tailored to designing shaping filters driven by white noise and to routine applications such as a predictive tool. © 2009.