Abstract: Abstract: Engineering accumulations are usually poor in soil structure and susceptible to erosion and corrosion, especially when the slope is steep. These could endanger their ecological functions and result in environmental hazards. How to improve ecological functions and alleviate soil erosion of earth slopes is pressing. The objective of this paper is to experimentally study the efficacy of different vegetation patterns in regulating water and sediment erosion of slopes under different water flow conditions. The experiment was conducted in the field at the Linghou Experimental Station, Institute of Soil and Water Conservation, Chinese Academy of Sciences (34°19′24″ N, 107°59′36″E). In the experiment, we analyzed the efficacy of five grass coverage patterns: upper coverage, middle coverage, low slope, band, and no coverage under four upslope runoff patterns (constant, earlier peak, medium peak, later peak). The experimental slope was a 20 m×1 m plot 0.5 m thick excavated from a slop; The slope angle was 32o. We used the fractal dimension of the eroded sediment particles to quantify the change in soil erosion, the enrichment rate (ER) to describe the difficulty associated with migration of eroded particles of certain size group, and the agglomeration rate (AR) to describe the ways that the surface runoff carrying the eroded particles of certain size group. The results showed that: 1) The relationship between soil denudation rate and the runoff power follows power function model (R2=0.47-0.72，P<0.01), and that the runoff power was the optimal parameter to describe the erosion dynamics of the slope. 2) Runoff can be reduced by 12.23%-49.62%, and sediment can be decreased by 12.92%-80.54% under vegetation patterns, the sediment reduction effect of vegetation pattern was higher than runoff reduction. The average runoff and sediment reduction benefit of the band and upslope patterns was 43.87%, 58.09% and 30.55%, 54.41%, respectively, which were significantly better than other vegetation patterns. 3) Compared with the control plot (without grass coverage), the volume percentage of sand particle in sediment of the vegetation plots was reduced by 18.79%-35.80%, the volume percentage of clay was increased by 3.56%-10.69%, which indicated that the interception effect of vegetation on sand was significant. Volumetric fractal dimension of the eroded particles was dominated by the volumetric fraction of clay and they were linearly correlated at significant level (R2=0.90，P<0.01). 4). Compared with the control, grass coverage reduced the sand enrichment rate, but increased the clay enrichment rate; their associated volumetric fractal dimension also increased. Clay and sand in the eroded sediments were in the forms of agglomerates, while the silts moved mainly as single particles. Our study has an important implication for ameliorating soil erosion from engineering accumulation.