Abstract: Soil permeability was one of the key factors affecting rainwater infiltration, and also an important prerequisite for analyzing the ability of soil to storage rainwater. In view of the current situation that the measured data of urban green space soil permeability was missing, the evaluation method was rare, and the index was insufficient, this paper obtained the data of soil permeable rate, void ratio and moisture content of different urban functional areas by using the actual measurement method. Correlation analysis, factor analysis and multiple linear regression analysis were used to analyze the main factors affecting the soil permeability of urban communities, and exploded to establish a soil permeability evaluation model with higher precision, less data demand and shorter calculation time. This model was used to obtain the infiltration capacity of urban green space rainwater in Beijing, establish the relationship between soil and rainfall-runoff, and get the measured data-oriented, closer to the actual runoff coefficient. The results showed that the soil permeability of the three functional areas in experimental plots were in order of residential area > cultural and educational area > commercial area, and the measured infiltration curves were all consistent with the Horton model. The void ratio, initial moisture content and initial infiltration rate of the internal factors of the soil were closely related to each other. The degree of vegetation coverage in the external factors had a strong negative correlation with human disturbance. Considering the combined effect of internal and external factors, the initial infiltration rate could establish a preliminary linear relationship with initial moisture content and vegetation coverage (R²>0.9). As the proportion of rainfall and impervious area ratio increased, the infiltration amount had smaller impact on runoff coefficient. The ratio of infiltration to rainfall corresponding to the total annual runoff control rate was between 37.6% and 58.2%, and it was negatively correlated with rainfall. The research could provide a theoretical reference for the modeling of sponge city and the design optimization of underlying surface.