This paper presents an erosion model, ARMOUR, which simulates time-varying runoff, erosion, deposition and surface armour evolution down a hillslope either as a result of a single erosion event or as the cumulative impact of many events over periods up to decades. ARMOUR simulates sediment transport for both cohesive and non-cohesive soil and dynamically differentiates between 'transport-limited' and 'source-limited' processes. A variety of feasible processes for entrainment of different size classes can be modelled and evaluated against data. The generalized likelihood of uncertainty estimation (GLUE) technique was used to calibrate and validate ARMOUR using data collected during rainfall simulator experiments at two contrasting sites: (1) non-cohesive stony sediments at Ranger Uranium Mine, Northern Territory, Australia; and (2) cohesive silty sediments at Northparkes Gold Mine, NSW, Australia. The spatial and temporal variations of model predictions within the individual runoff events showed that some entrainment processes could not model the spikes in concentration and subsequent depletion, while the hiding model of Andrews and Parker best simulated the concentration trends for both calibrated and independent runoff events. ARMOUR also successfully captured the coarsening of the surface material, though small, over the duration of the rainfall simulator trials. This was driven by the depletion of the finest size class of the soil. For a constant discharge, ARMOUR simulated higher sediment flux at the start of the storm with the sediment flux and concentration diminishing with time. For natural rainfall a power law relationship between sediment flux and discharge was observed. The calibration exercise showed that sediment concentration and discharge alone are insufficient to calibrate all aspects of the physics, in particular the armour depth. This appears to be because the armouring during the short duration events is driven by depletion of the finest classes of the sediments (diameters less then 62·5 µ µ µ µ µm), which are not normally measured.