The characterization and modeling of wind conditions have received an increasing attention over the past decades. Indeed they influence many human activities (renewable energy, shipping, aviation, ...). Wind conditions are measured with various instruments, wind speed is generally collected as averages over a given time-window. However, time-averaged wind speed might not be representative of potential sub-sample variability as for instance wind extremes. We propose to investigate statistically the discrepancy between the available measurements of wind speed as time-averages across different time-scales and characterize the sub-sample variability arising from across scales through the use of multi-resolution statistical methods.