Effect of DEM Resolution on Soil Erosion Estimation and Gully Erosion Susceptibility Prediction

In the present study, we integrated detailed field investigations with high-resolution remote sensing products to investigate the effect of digital elevation model (DEM) resolution on estimating soil erosion and predicting gully erosion susceptibility in six representative watersheds across contrasting (highland, midland, and lowland) agro-ecological environments in the Upper Blue Nile basin of Ethiopia. Each agro-ecological environment comprises paired-watersheds relatively under different soil and water conservation treatment conditions. High-resolution Enhanced DEMs were aggregated in a geographic information system environment to produce eight datasets with varying spatial resolutions ranging from 0.5 to 30 m. The revised universal soil loss equation (RUSLE) was used to estimate soil erosion while the random forest (RF) was used to predict the severity of gully erosion by randomly selecting about 70% and 30% of the data in each gully dataset for training and validation purposes, respectively. The topographic characteristics of watersheds related to elevation, aspect, and size of watersheds were identified to be influenced by the DEM resolution in addition to the slope-length (L), slope-steepness (S), and length-slope (LS) factors of RUSLE. The separate analysis of L and S factors showed overestimations and underestimations, respectively, compared to the finest pixel resolutions when aggregation continued towards coarser resolutions while the combined LS factor-based analysis showed an increase of up to 29% and a decrease of up to 77% in soil loss estimations. On the other hand, gully erosion susceptibility prediction improved with an increase in spatial resolution, and RF showed outstanding performance (Kappa > 96%, sensitivity > 97%, specificity > 99%, and overall accuracy > 98%) when the finest-resolution datasets were used. The spatial resolution of DEM was found to have an impact on the topographic attributes of watersheds, soil loss estimations, and gully erosion severity predictions. Therefore, studies focusing on the topographic and geomorphic Earth surface processes in watersheds need to consider using the finest available spatial resolution of DEMs to get accurate estimations and predictions.