In order to solve the water shortage, uncoordinated runoff-sediment relationship, and prevent carbon loss in the Yellow River Basin, there is an urgent need to explore the rational vegetation configuration in order to relative increase runoff, reduce sediment and prevent soil organic carbon loss in the gully-slopes on the Loess Plateau. To solve these problems, the simulated rainfall experiments with an intensity of 60 mm·h−1 were conducted in four types of plots: shrub-grass (SG), tap root grasses (GT), fibrous root grasses (GF) and bare land (CK). Four types of plots were built on one gully-slope with a slope of 35°–40° in situ. Compared with CK (mean runoff rate: 0.61 mm·min−1), the runoff reduction benefits of SG (0.27 mm·min−1), GT (0.04 mm·min−1) and GF (0.17 mm·min−1) plots were 54.9 %, 71.7 %, 92.2 %. Compared with GT and GF plots, the total runoff of SG plots was 1.6 to 5.8 times. Compared with CK, the sediment reduction benefits of three vegetation plots were more than 98 %. Shrub-grass is a more suitable vegetation combination for soil and water conservation: reducing sediment and relative increasing runoff. The soil organic carbon loss has been well alleviated through vegetation restoration, and vegetation plots all prevented over 98 % of soil organic carbon loss compared with the bare plot. Soil organic carbon loss equation based on runoff and sediment was established and validated. The R2, P and NSE results of the equations were 0.92–0.95, <0.001, 0.85–0.86, respectively, indicating that the fitted equations can be considered acceptable. The soil organic carbon loss in the gully-slopes was mainly from sediment (more than 85 %) compared with runoff. Therefore, controlling sediment is more effective than controlling runoff in preventing soil organic carbon loss in the gully-slopes on the Loess Plateau. These results can provide a basis for decision-making of high-quality development of contemporary soil and water conservation on the Loess Plateau.