We report on a series of soil erosion experiments performed on the 2-m x 6-m EPFL erosion flume. Total sediment concentrations and the concentrations of seven size fractions (<2, 2–20, 20–50, 50–100, 100–315, 315– 1000, and >1000 mm) were measured during 14 high- intensity (47.5–52.5 mm/h) rainfall experiments on three different slopes (2.2– 12.4%). The short-time and long- time analytical solutions for the Hairsine-Rose erosion model were rewritten to account for infiltration. The newly collected data were used to test the model under conditions that had not been explored before (steeper slopes, infiltration, more realistic soil composition). The analytical solutions could predict the observed total sediment concentration well. However, the observed sediment concentrations for the individual size classes could be predicted only when adjusted settling velocities were used. The adjusted settling velocities were estimated through manual optimization. The optimized settling velocities for the smallest and midsize particles (<100 mm) were larger than calculated from Stokes’ law or measured in a 0.47-m tube, while the optimized settling velocities for the largest particles (> 315 mm) were smaller than measured. The effective settling velocities could also be calculated from the calculated amount of material in the shield of each size class at the end of the experiments. This calculated settling velocity distribution agreed very well with the optimized settling velocities. This study moves the Hairsine-Rose model another step closer to an operational soil erosion model for field applications.