Key points are not available for this paper at this time.
Several environmental factors are known to influence the spatial distribution and susceptibility of gully erosion, yet the relative importance and interaction of these factors remain little understood in Ethiopia. In this study, we integrated detailed field investigations with high-resolution remote sensing products to assess gully erosion susceptibility and identify its controlling factors using Random Forest (RF) model in six representative watersheds across contrasting (highland, midland, and lowland) agro-ecological environments in the Upper Blue Nile basin of Ethiopia. Data for 20 controlling factors were extracted from datasets at eight different pixel resolutions ranging from 0.5 to 30 m in a geographic information system environment. About 70 % and 30 % of the dataset in each watershed were randomly selected for model training and validation purposes, respectively. Multicollinearity and correlation analyses were performed to identify variables with collinearity problems and explain their statistical relationships among the other variables. RF predicted gully erosion susceptibility and the relative importance of the controlling factors. The model showed outstanding performance when the finest-resolution datasets were used. Elevation, height above nearest drainage, runoff curve number-II, distance from streams, drainage density, soil type, and land use/land cover were found to be the most important factors controlling the spatial distribution of gullies in all six watersheds, irrespective of the watershed treatment conditions and agro-ecological settings. Thus, the most susceptible land to gully erosion was low-lying grazing and cultivated lands with sensitive soil of high runoff-generation capacity located within short horizontal and vertical distances from drainage networks. Therefore, basin- and watershed-scale gully management strategies should give priority to these areas. The identification of the runoff curve number-II, a hydrologic parameter for predicting direct runoff from excess rainfall, as one of the most important factors controlling gully erosion was a novel finding which will be useful for developing improved process-based gully erosion models.
Setargie et al. (Mon,) studied this question.