Accumulation has become the main source of soil erosion in the production and construction projects. Soil detachment is a key process affecting soil erosion since it determines the amount of sediment that is potentially transferred to surface water bodies. Soil detachment capacity is a key parameter in many process based erosion models such as the Water Erosion Prediction Project model. Therefore, quantifying detachment capacity of the engineering accumulation under different conditions is pivotal to calibrate and validate the process based on the erosion models. In the study, we used the soil detachment capacity to quantify the variation ranges of the engineering accumulation in the different area, and evaluate the regulation of vegetation restoration patterns, years of recovery and the slope cutting and grading control on the soil anti-scour erodibility. Different engineering accumulations of the expressways in the Qin-ba mountain area, Guanzhong Plain and Hilly region of Loess Plateau (in Shaanxi Province) were selected. Undisturbed soil samples were collected from the surface soil using the cutting rings with a diameter of 10 cm and a height of 6.37 cm for the measurement of soil detachment capacity. The soil detachment capacity was measured in a 2.0 m long and 0.15 m wide hydraulic flume indoors. Flow discharge was controlled by six valves and measured 5 times with plastic buckets and a volumetric cylinder. After the flow became stable, the flow surface velocity was measured using a fluorescent dye technique for 10 times. The velocity was multiplied by a reduction factor of 0.67 to obtain the mean flow velocity. The results indicated that the variation ranges of the soil detachment capacity on typical accumulation in the Qinba mountainous area, Guanzhong Plain and Hilly area of Loess Plateau were 0.034-1.659, 0.311-0.816 and 0.346-1.042 kg/(m2·s), respectively. Compared with agropyron cristatum, the natural restoration of vegetation on the slope of the engineering accumulation with coronilla varia can significantly reduce the soil detachment capacity, and the reduction was 94.97%. Compared with no-reclamation, there was no significant effect on soil detachment capacity of the corn and beans on the slope of accumulation which was constituted with rock fragments and soil in the short term. Compared with recovering in one year, there were no significant (P>0.05) differences in recovering two years, but the soil detachment capacity was remarkably decreased by recovering five years, and the reduction was 57.35%. Furthermore, compared with the cultivated land, the reduction of soil detachment capacity was 60.41% in recovering five years. There was no significant difference in the spatial variation of soil detachment capacity for accumulation with Heilu soil in the short slope (the length of slope was less than 60 m). Compared with untreated slope, the soil detachment capacity was remarkably decreased by the slope cutting and grading control on the engineering accumulation. And compared with the untreated slope, the reduction of soil detachment capacity in the platform and slope by the regulated measures were 66.79% and 49.04%, respectively. There was significant negative (P<0.05) correlation between root mass density, soil cohesion, soil water content, median soil grain size, the clay content and the soil detachment capacity, using an exponential function. The prediction model of soil detachment capacity was well fitted with on root mass density and flow shear stress. This research can not only provide the basic data for the prediction soil detachment capacity of accumulation, but also provide guidance for the treatment measures in the slope of engineering accumulation.