Societies can be better prepared to face hydrological extremes (e.g. flash floods) by understanding the trends and variability of rainfall aggressiveness and its derivative, erosivity density (ED). Estimating extended time series of ED is, however, scientifically challenging because of the paucity of long-term high-resolution pluviometric observations. This research presents the longest ED time series reconstruction (1701-2019) in northwest Italy (Piedmont region) to date, which is analysed to identify damaging hydrological periods. With this aim, we developed a model consistent with a sample (1981-2015) of detailed novel Revised Universal Soil Loss Erosion-based high-resolution data and documentary hydrological extreme records. The modelled data show a noticeable rising trend in ED from 1897 onwards, together with an increase of extreme values for return periods of 10 and 50 years, consistent with the Clausius-Clapeyron scaling of extreme rainfall. We also suggest the North Atlantic Oscillation and Atlantic Multidecadal Oscillation may be associated with rainfall extremes in Piedmont.