Accurate prediction of long-term ‘characteristic’ loads associated with an ultimate limit state for design of a 5-MW bottom-supported offshore wind turbine is the focus of this study. Specifically, we focus on predicting the long-term fore–aft tower bending moment at the mudline and the out-of-plane bending moment at the blade root of a monopile-supported shallow-water offshore wind turbine. We employ alternative probabilistic predictions of long-term loads using inverse reliability procedures in establishing the characteristic loads for design. Because load variability depends on the environmental conditions (defining the wind speed and wave height), we show that long-term predictions that explicitly account for such load variability are more accurate, especially for environmental states associated with above-rated wind speeds and associated wave heights. Copyright © 2012 John Wiley & Sons, Ltd.