Assessing the relative erodibility of someKenyan soils using a rainfall simulator and the prediction of relative erodibility
This study was an attempt to assess the relative erodibility of some Kenyan soils using a rainfall simulator, and to develop a regression equation based on easily measured soil properties that would predict relative soil erodibilities. The surface horizons of 15 contrasting soils from cultivated land were selected for this study. The air-dried soils were packed into metal trays and positioned on a 6° incline beneath a rotating disc rainfall simulator. The duration of the simulated (47mm/hr) rainstorms was lhr, except for 3 soils where a l^hr storm was necessary in order to obtain runoff. 2 The soil losses ranged from 2.03 to 7.91g/m / unit of erosivity, while the relative erodibility k factors ranged from 0.054 to 0.210. For five of these soils, the soil losses correlated fairly well (r=0.79) with measured soil losses obtained under field conditions using the same rainfall simulator and similar rainfall intensities. The k values for all 15 soils correlated rather poorly (r=0.54*) with the erodibility factors K obta ned from Wischmeier's soil erodibility nomograph, but the k factors of the 13 non-swelling, well drained soils (n) correlated reasonably well (r=0.84***) with Knom. Spearman's ranking coefficients for the k factors as given by Knom values for all 15 and for the 13 soils were rs=0.679** and rs=0.885** respectively. To try and improve on the prediction of the k values, simple and multiple linear regression analyses on the k factors of the 13 non-swelling, well drained soils were carried out using various easily measured soil properties. The best single predictive factors for the relative erodibilities were the % fine sand (r=0.79**), % clay (r=-0.62*), % organic matter (r=-0.54), % fine and very fine sand (r=0.S0***) and dispersion ratio (r=0.75**). In a multiple linear regression analysis, four soil properties viz. dispersion ratio, % clay, % organic matter and bulk density explained 90% of the variations in k. In comparison, Wischmeier's soil erodibility nomograph accounted for only 71% of the variations in k for the 13 soils though both Wischmeier's relationship and the multiple regression were able to rank the soils equally well in order of their relative erodibility factors (rs=0.88** and 0.87** respectively).