Time varying mean extraction for stationary and nonstationary winds.

This paper discusses different strategies for the extraction of the time-varying mean from wind speed time histories. Due to the advantage of allowing analytical evaluations, the attention is focused on kernel regression techniques, considering different weighting functions, namely a constant, a Gaussian and a cardinal sine weighting function. The problem is firstly treated analytically, and the frequency-domain properties of the filter associated to different kinds of weighting functions in the definition of the slowly varying mean through kernel regression are analysed. Then, different weighting functions are adopted for the analysis of digitally-simulated stationary wind speed time histories and for the time histories of thunderstorm outflows recorded by a tri-axial anemometer. The consequences of the adoption of different weighting functions on the harmonic content and statistical properties of turbulence are studied. The same features are found also for thunderstorm outflow records. • Kernel regression techniques for the time-varying mean extraction are studied. • The properties of the filter associated to various weighting functions are analysed. • The statistical moments of the reduced turbulent fluctuation are estimated. • Applications to stationary and nonstationary wind speed time histories are provided. [ABSTRACT FROM AUTHOR]