This paper investigates the transient characteristics in 41 velocity records of 19 thunderstorm events from around the world—9 from Europe, 9 from the United States, and 1 from Australia. The transient features of thunderstorm winds were examined by introducing an objective method for the detection of changepoints in the time series. The methodology divides velocity records into segments characterized by different statistical properties. The segmentation is based on the following properties of the isolated segments: mean (M) and the standard deviation (SD). This study demonstrated that the maximum velocity during the thunderstorm peak in the events from Europe is typically 2–4 times larger than the mean wind speed before the thunderstorm. The duration of the thunderstorm velocity peak was 2–5 min in approximately 60% of the analyzed records using the M statistic and 5–10 min when analyzed using the SD statistic. Therefore, the velocity fluctuations caused by thunderstorm winds last longer than the abrupt changes in the mean wind field. Similarly, the ramp-up time was longer when the records were analyzed using the SD statistic. The segmentation methodology was tested for different duration of velocity records and using data with different sampling frequencies. The performances of the introduced method were compared against the results of two other segmentation procedures proposed in the literature. One of the practical applications of this method is the physical separation between the thunderstorm and non-thunderstorm components of a wind event.