...
The choice of a 50 ms-1 cut off is an arbitrary one, which could miss spurious values that are less than 50 ms-1. It is also possible, that this cut off point could penalise realistic winds, such as in hurricanes. However, given the spatial resolution of ERA5 data, it is thought that in general, 10m wind components higher than 50 ms-1 are not realistic, even if the observed magnitude is higher - this has been seen in at least one the case of a hurricane on 12 September 2004 at 20 UTC, when the 10 metre V wind component was -53 ms-1, see Table 2, below.
The tables below, for January 1979 to 18 February 2020, give details when the 10m wind components (the 10 metre U wind component and 10 metre V wind component) exceed 50 ms-1.
...