Hydrological model is an effective tool for hydrological research. The initial abstraction ratio (λ) is a key parameter of SCS-CN model, a commonly used runoff model of great significance to simulate the hydrological process at the watershed scale. In order to examine the effects of biological soil crusts (biocrusts) on λ and improve the accuracy of the model used in the restored grasslands where biocrusts widely presented in the Loess Plateau region, we firstly determined the relationship between the amount of the potential maximum infiltration (S) and the amount of the actual infiltration (F), and then investigated the effects of biocrust coverage on λ by using the simulated rainfall experiment in the Yingwoshanjian watershed in Dingbian County, Shaanxi Province. The revised model was verified by the runoff results of the simulated rainfall experiments in the Zhifanggou watershed in Ansai County, Shaanxi Province. The results showed that the relationship between S and F on biocrust slope was described as S/F=2.5×60/T (where T was the rainfall duration). There was a negative correlation between λ and biocrust coverage (CBSC) described as λ=0.0791×e(-0.015×CBSC), R2=0.60. Compared with that using the standard value of λ, the efficiency coefficient of the model was increased by 338.7% and the qualified rate was increased by 16.1% after revising λ according to the biocrust coverage. The results provided a scientific basis for the calibration of λ on biocrust slopes in the Loess Plateau region, and were of great significance to accurately assess the hydrological effects of the implementation of the "Grain for Green" Program on the Loess Plateau.