In the first paper in this series, we demonstrated that most process-based erosion models have a series of in-built assumptions that led us to question their true process basis. An alternative soil-erosion model (MAHLERAN – Model for Assessing Hillslope-Landscape Erosion, Runoff And Nutrients) based upon particle-travel distance has been presented in the first paper in this series and this paper presents the first of two evaluations of the model. Here, a sensitivity analysis shows that the numerical model is consistent with the analytical model of Parsons *et al.* (2004) and demonstrates that downslope patterns of sediment flux on hillslopes are a complex interaction of rainfall intensity, duration and pattern; hillslope gradient; surface roughness and sediment size. This result indicates that the spatial scaling of sediment transfers on hillslopes is a non-trivial problem and will vary from point to point and from event to event and thus from year to year. The model is evaluated against field data from a rainfall-simulation experiment on an 18 m × 35 m plot for which there are sub-plot-scale data on runoff hydraulics and sediment flux. The results show that the model is capable of reproducing the sedigraph with an overall normalized root-mean-square error of 18·4% and Nash–Sutcliffe efficiency of 0·90. Spatial and temporal patterns of particle-size distributions of the eroded sediment are also reproduced very well, once erosion parameters have been optimized for the specific soil conditions. Copyright © 2008 John Wiley & Sons, Ltd.