The applicability of SCS-CN model developed by Soil Conservation Service of United States Department of Agriculture (USDA) is still disputed. The calculation of curve number (CN) values in other countries or regions using assigned fixed initial abstraction ratio is inaccurate. It is necessary for the further calibration of parameters, especially when the model runs in the areas with complex natural environment, such as the Chinese Loess Plateau. In this study, based on the observed data of 47 runoff plots for five years during the growing season, the CN values of typical terracing and vegetation measures were determined by the arithmetic mean method, and the empirical equation method was used to identify the saturated hydraulic conductivity and hydrologic soil groups. The results indicate that ① the soil in middle region of Gansu Province is made up of loam and slity loam, whose saturated hydraulic conductivity is between 18 and 180 mm/h, and the minimum infiltration rate is between 3.81 and 7.26 mm/h. ② The CN was strongly affected by antecedent soil moisture, plant species, vegetation cover, land use, slope and terracing measures. There is an obvious difference between the calculation with correction in terms of slope and without. ③ The CN₁ values in different typical vegetation patterns sort by size:shrubland (Hippophae rhamnoides) < abandoned land (Agropyron cristatum) < artificial grassland (Medicago Sativa) < farmland (Triticum aestivum) < arboreal land (Pinus tabulaeformis), and the CN values decrease with increasing vegetation coverage among the same kind of plots with same slope gradient. ④ The terracing measures including zig terraces, leveled benches and leveled ditches all increased the saturated hydraulic conductivity and decreased the CN values in different antecedent moisture conditions remarkably.