How Strong Is the 50-Year Wind Gust? Annual Maxima of 5-Second Gusts at 20 US Airports, 1996-2025, a Closed-Form Error for the Gumbel Return Level, and a Rising Gust Factor That Looks Like a Measurement Change

Design wind speeds for buildings come from extreme value fits to records of peak gusts. I fit Gumbel, generalised extreme value (GEV) and peaks-over-threshold models to the daily fastest 5-second gust (WSF5) from GHCN-Daily at 20 large US airports, 1996 to 2025, after removing 130 of 215,488 station-days that fail simple physical checks, including a recorded gust of 102.4 m/s at Duluth. For the maximum likelihood Gumbel fit I prove that the large-sample variance of the T-year return level is (sigma squared over n) times 1 + (6/pi squared)(y_T + 1 - gamma) squared, where y_T is the reduced variate and gamma is Euler's constant. With 29 annual maxima and a median scale of 2.67 m/s, the 50-year gust has a standard error of about 1.7 m/s, and the 700-year gust of 2.7 m/s. The median 50-year gust is 35.1 m/s (range 27.4 to 40.0) and the median 700-year gust 41.9 m/s (93.7 mph). The GEV shape is negative at 12 of 20 stations, and the Gumbel model is rejected at 3. Annual maximum gusts rose at all but one station between 1996-2010 and 2011-2025, by a median 1.87 m/s, with significant upward trends at 10 stations. Over the same years the daily average wind speed did not rise, while the ratio of mean gust to mean wind speed rose significantly at 19 of 20 stations. This pattern points to a change in how gusts are measured or processed, not a windier climate. I show that a step of this size biases a stationary 50-year estimate by about 0.6 m/s.