The folly of using extreme-value methods in full-scale experiments

Extreme-value methods have been applied successfully to meteorological measurements of wind speed to predict extreme events which have a given probability of exceedence. Attempts have been made to extend the method to give extreme values of loads on buildings. Although this method may have application to stationary flows, it is demonstrated here that it is misleading to apply extreme-value methods to full-scale wind-load measurements where the flow is intrinsically non-stationary. This finding may also apply to many wind-tunnel studies where records are of insufficient length to satisfy the condition for stationarity or the flow properties are not controlled adequately.