The purpose of this study was to characterize the soil organic carbon dynamics associated with four land-uses (cropland, grassland, shrubland, and forestland) in the upper Blue Nile Basin of the Ethiopian Highlands. We collected diverse biophysical data to allow spatial variability of soil organic carbon and factors contributing to this variation to be determined statistically, and used well established simulation models to interpret the data, predict long-term carbon dynamics and determine the potential for improvements in soil quality and mitigation of greenhouse gas emissions. The spatial variation in soil organic carbon in the 0-20 cm soil depth was significant across study areas (P<0.01, 21.6 g kg-1-37.8 g kg-1), and between land-uses (P<0.05, 27.9 g kg-1 in cropland and 43.0 g kg-1 soil in forestland). In a multiple linear regression model, among the 11 explanatory variables used, four (total nitrogen, shrubs, trees and land use) showed a significant positive effect (P<0.01), while three (impact of grazing, impact of erosion and clay) showed a significant negative effect (P<0.01). Simulations using the assumption of steady state to estimate carbon dynamics suggested that plant inputs from croplands are generally lower than grass or shrublands, resulting in build-up of recalcitrant organic matter in shrublands compared to grass or croplands. Erosion results in a decline in both the absolute and relative amounts of carbon in recalcitrant pools, but the more active pools remained unchanged, meaning that the overall activity of soil organic matter is increased when the soil is eroded, and it becomes more difficult to sequester soil carbon. More rapid decrease in soil organic carbon is likely to be due to increased erosion, persistent removal of organic materials, grazing pressures, and higher rate of decompositions in cropland and grazing land-uses. Conversion to shrubland shows high potential to quickly restore eroded soils and build up a pool of recalcitrant soil carbon, suggesting that management of land using periodic 20 year exclosures, preventing cropping and grazing and allowing shrubland succession, could be beneficial in restoring degraded soils. Simulations using a calibration approach, rather than assuming steady state, corroborated these findings, suggesting that after 30 years of current management, croplands will deplete soil organic carbon by 5.6, 7.1 and 7.2 tha-1, and grasslands by 3.5, 3.9 and 2.7 tha-1, while shrublands will build-up soil organic carbon by 0.6, 0.1 and 1.3 tha-1 at study sites. At one study site, forests, will further increase soil organic carbon by 6.7 tha-1 after 30 years. The significant positive impact of shrubs and trees on soil organic carbon suggests the need to focus on introduction of agroforestry systems in crop and grasslands. © 2015 Elsevier B.V.