Hydrological extremes are major weather related disasters, but little is known about their long-term patterns in the context of environmental change. Better understanding of damaging rainfall (e.g. rainfall-erosivity events) occurring at different time-scales has important implications for hydrological and land degradation management. The study of the interdecadal variations may help in understanding some of the consequences of abrupt environmental changes over long time periods. Thus, a decadal-scale rainfall erosivity model (DREM), comparable with the Revised Universal Soil Loss Equation (RUSLE), was developed based on a parsimonious interpretation of rain aggressiveness (95th percentile of rainfalls). The DREM was parameterised to capture interdecadal erosivity variability at the Ukkel station (Belgium), which has the longest RUSLE-based rain-erosivity series in Europe (1898–2007). The DREM performed well against decadal RUSLE data, with a coefficient of determination (R2) of 0·72 and a Nash–Sutcliffe efficiency index of 0·71. The model outperformed three well-established models used in this study (R2 ~ 0·4). For a spatial evaluation of the DREM, a pattern of decadal rainfall erosivity was provided for an area around Ukkel, which includes the western part of Germany bordering Belgium, and was compared with maps from the RUSLE approach for 1961–1990. The 95th percentile of June–September rainfalls proved to be a better predictor of decadal rainfall erosivity than yearly based precipitation amount. These results lay the foundation for estimating decadal erosivity in the surrounding areas of Ukkle as well as for historical reconstructions where detailed hydrological data are unavailable, and assumptions cannot be met, for physically based models. Copyright © 2012 John Wiley & Sons, Ltd.