 A technique is presented for assessing the predictive uncertainty of rainfall runoff and hydraulic forecasts. The technique conditions forecast uncertainty on the forecasted value itself, based on retrospective quantile regression of hind-casted water level forecasts and forecast errors. To test the robustness of the method, a number of retrospective forecasts for different catchments across England and Wales have been used to derive in a probabilistic sense. This article was authored by A. H. Weertz, H. C. Wincemias and J. S. Vercade.