| Abstract: | Prompt location of areas exposed to high erosion is of the utmost importance for soil and water conservation planning. Erosion models can be useful tools to locate sources of sediment and areas of deposition within a catchment, but the reliability of model predictions of spatial patterns of erosion at catchment scale has seldom been validated against observations. This study aimed to evaluate the performance of a simple empirical model (Morgan, Morgan and Finney model, MMF) in predicting spatial patterns of erosion at two small catchments in the East African Highlands: Kwalei (Tanzania) and Gikuuri (Kenya). Erosion maps predicted by the MMF model were compared with erosion maps obtained by direct survey. In Kwalei, erosion features were especially frequent in fields of annual crops. In Gikuuri, slope was the critical erosion factor, with estimated erosion rates >10 kg m?2 a?1 on slopes >18 per cent. Predicted erosion rates were mainly transport‐limited and ranged from <0·01 to 13·50 kg m?2 a?1 in Kwalei and 9·29 kg m?2 a?1 in Gikuuri. The performance of the MMF model in predicting the spatial patterns of erosion was acceptable in Kwalei, but poor in Gikuuri. However, by excluding the elements at the valley bottoms in Gikuuri Catchment, the performance of the model improved dramatically. The spatial pattern of erosion predicted by the MMF model was driven by the accumulation of surface runoff, which did not consider the possibility of re‐infiltration along the slope. As a result, the MMF erosion patterns predicted by the model increased invariably from the ridges to the valley bottoms, hampering the model suitability for locating areas subjected to high and very high erosion. It is concluded that the model predictions could be substantially improved by introducing a more realistic hydrological component for the prediction of surface runoff along the hill‐slope. Copyright © 2005 John Wiley & Sons, Ltd. |
