The results contained within this paper confirm the extreme complexity of erosion processes, especially relationships between soil loss and rainfall intensity. In this study this relationship was statistically low (r2 = 0.368). This increased significantly when other factors were also considered: Rum de la USLE-M (runoff ratio, rainfall intensity and storm kinetic energy), soil resistance to drop detachment, slope angle and gravel cover. Incorporating these factors to a multiple regression analysis, a r2 of 0.74 was calculated. In the equation the gravel cover and Rum were the most significant variables with a beta weight (partial regression coefficient) of −0.820 and 0.744, respectively. Gravels on the soil surface protect it against erosion. A stony soil favours more rapid infiltration, diminishes runoff discharge, absorbs the kinetic energy of raindrops and dissipates the overland flow. These issues have been considered by Poesen and Lavee (1994). A layer of rock fragments has been used in some regions to protect the soil beneath it from erosion and to increase the quantity and quality of grapes ( Natchtergaele et al., 1998).
The equation of multiple regression analysis predicts soil losses with a great certainty. However, an over estimation is observed in events of low erosion and rainfall intensity. The deviation could be explained by the control of the water storage in depressions of soil surface and the low energy of the rainfall drops in simulation tests of low intensity. The over estimation is much more important in the relation between the results of our equation and the obtained ones with the USLE-M. The efficiency coefficient is low and negative. This could be related to the factor LS that incorporates the USLE-M. This factor has little relevance in Eq. (9) since the length has not been incorporated and the slope has little influence in the multiple regression analysis (beta weight: 0.148). It is also necessary to emphasize the control of gravel cover in our equation to explain low erosion rates.