The experiments described in Salles and Poesen (2000) focused on the detachment of soil particles from a bare soil surface by raindrop impact and their subsequent transport by raindrop splash. The selected experimental set-up, i.e. simulated rain and the use of splash cups, allows the measurement of a combination of both processes involved in the soil movement during raindrop impact: i.e. detachment by raindrop impact and transport by drop splash outside the splash cup. It would be almost impossible to differentiate between the two erosion subprocesses. As Hillel (1998) noted "The abrupt collision of drops with the soil has the impact and effect of a miniature bomb: it detaches and splashes soil particles".
In Kinnell's (1982) study, the simulated rain was made up of drops, all having the same drop size. In such experimental conditions rain intensity (I ) will be proportional to D 3 , if D is the unique drop diameter of the rain. The square of the mass of an individual drop is proportional to D 6 . Therefore, the square of the drop mass and I 2 must be related. Attention should be paid to the fact that this relation is valid only if the drop size is unique, as was the case in these experimental conditions, which were quite different from those found in natural rain.
The experimental procedure in Salles and Poesen (2000) was not chosen randomly. We mentioned that before the splash cup was exposed to the rain "the surface of the sand was made flush with the rim of the splash cup". In so doing, the sediment losses from the cup by drop impacts pushing sediment sideways over the edge of the container (as observed by Kinnell, 1974) were minimized during the experiments. We are, however, aware of the fact that 'minimized' does not mean 'negligible'.
The other comment concerns the variation of splashed sediment with time. The duration of the rainfall application during our experiments was a compromise between a relevant mass of detached sediment (that could be measured) and a minimum lowering of the soil surface in the splash cup. We did not check systematically if the rate of soil loss varied with the duration of rain exposure for all tested rain intensities. Nevertheless, for some of the tested rain intensities, the duration of rain exposure varied between replicates. No discrepancies were found between the rates of soil losses derived from these replicates, and therefore dependence of splash losses on the duration of the rain exposure could not be demonstrated with the rain exposure duration tested.
These answers provide more background on the experimental conditions used and on the assumptions made during the analysis of the experimental data. As already mentioned by Kinnell (1974Kinnell ( , 1982) ) and Poesen and Torri (1988), the splash cup technique is far from perfect. Nevertheless, experiments conducted with many replicates and satisfying some basic conditions allowed the identification and interpretation of the role of the erosion factor rain in the raindrop impact and splash detachment process.
At least, it appears obvious that the erodibility of a soil will change with rain intensity applied because of the evolution of the soil surface state during rain exposure. This point raises the following question: when