Tillage‐induced surface roughness and topographic conditions for rill initiation in Angereb watershed, Ethiopia

Abstract

AbstractRill erosion makes up one source of sediment detachment and transport on agricultural fields in the Ethiopia highlands. It is so important to understand the conditions for rill initiation and demarcating rill prone areas for management planning. In this paper, we estimated rill incision considering tillage‐induced surface roughness, upslope contributing area, and slope gradient. Field experiments were installed on two fresh‐tilled smallholder agricultural plots of 3.2 m wide and 14.4 m long in the Angereb watershed of the upper Lake Tana subbasin in Ethiopia. The plots mainly represent clay and sandy clay loam soils. Time series digital elevation models generated by digital photogrammetry were used to derive topographic and surface roughness variables corresponding to 46, 116, 174, and 410‐mm cumulative rainfall. A non‐linear regression model was fitted on the slope gradient, upslope contributing area, and surface random roughness of individual rills to predict threshold area for rill initiation. Subsequent rainfall events have resulted in rough surfaces later. The change in random roughness in each rainfall event compared with the first stage amounts to 1.81, 1.62, 1.68, and 1.74 cm on clay soil and −0.05, 0.86, 1.55, and 2.06 cm on sandy clay loam soil. Prediction of the threshold area for rill initiation has improved when considering random roughness and its temporal changes. The average prediction parameter of the unit contributing area was 0.03 ± 0.05 and 0.04 ± 0.04, and of the random roughness was −0.08 ± 0.13 and −0.13 ± 0.05 on clay and sandy clay loam soils, respectively. The prediction resulted in a threshold area between 0.20 and 0.47 (average of 0.36 ± 0.11) on clay soil and 0.28 and 0.39 (average of 0.34 ± 0.05) on sandy clay loam soil. The significant variation of threshold area (p < 0.05) over the subsequent rainfall events ensures the role of incorporating surface roughness in predicting rill initiation. The result demonstrates the need to consider the variable threshold area over rainfall periods for rill incision modelling and indirectly improves the slope–length factor of erosion models.