Accurately predicting farmland surface runoff is the basis of controlling agricultural non-point source pollution. According to the motion characteristics of rainfall and surface runoff, coupled with the Green-Ampt model, the apparent function of cumulative infiltration, the Kinematic wave model and the steady-state sediment yield model, an approximate semi-analytical model for slope runoff and sediment yield under rainfall conditions was established. The model approximated the Kinematic wave equation through linear expression of the relationship between the infiltration rate and the water depth. By setting the simulated rainfall test under five different rain intensities between 30~90 mm/h, the model parameters were derived and the model accuracy was verified, and the sensitivity of the model parameters were analyzed. The results showed that: (1) The calculated value of runoff and sediment model was in good agreement with the measured value, and the model was more suitable for simulating the runoff and sediment yield process under 60 mm/h rainfall intensity, and the Manning's roughness coefficient distributed between 0.000 26~0.000 29 min/cm^1/3, and the relationship between the infiltration rate parameter c and the rain intensities could be described by an exponential function c=9×10^-4p^-0.644. The parameters a and b in the sediment yield model distributed between 0.13~0.15 and 0.10~0.12, respectively. Due to the accidental factors of sediment yield, the model simulation accuracy of the sediment yield process was lower than that of the runoff process. When the value of c was 0~0.05, the value of Dif changed little, it indicated that the soil surface runoff was not sensitive to c. In the process of runoff sediment yield, the value of Dif varied with the change of parameters a and b. Overall, runoff sediment yield was more sensitive to the runoff erosion coefficient parameter a than the parameter raindrop erosion coefficient b.