In this paper, we employ the multiscale multifractal analysis (MMA) method to investigate the fractal properties of wind speed records depending on their magnitude of the fluctuations and the timescale. The MMA results show that the high-frequency wind speed records appear to be far more complex and contain abundant information, which cannot be detected by the popular scaling analysis method, i.e., multifractal detrended fluctuation analysis (MF-DFA). Comparing the Hurst surfaces of nine groups of wind speed data, we find that for the negative qs, all the surfaces exhibit intensive fluctuations and significant differences. In addition, the distribution histograms of Hurst surfaces for the positive qs reveal that the large fluctuations of all wind speed data depend on the spatial positions, which is further illustrated by the wind roses. Subsequent analysis of shuffled and surrogate series reveals that the multifractality of wind speed time series is mainly stemming from the long-range correlation, while has less to do with broad probability density function. Finally, the effect of sampling period is discussed. The results suggest that a sampling period of 20 min is sufficient to characterize multiscale multifractal properties of high-frequency wind speed data.

No abstract