Based on statistical theory, a runoff-yield model, considering the spatial variations of rainfall, soil infiltration capacity and water storage capacity, is proposed in this paper. It is supposed that the spatial variations of a rainfall event could be described using a Probability Density Function(PDF) or a Cumulative Distribution Function(CDF), and the specific PDF or CDF at every time step of the rainfall event is estimated by adopting the goodness-of-fit approach to match the curve with the real rainfall data. The parabolic types of mathematical functions are used to represent the spatial distributions of soil infiltration capacity and water storage capacity. According to the joint probability distribution of rainfall and soil infiltration capacity, the distribution of surface runoff is deduced from the infiltration excess mechanism, and the further analytical solution to surface runoff is obtained. Infiltration supplements soil moisture, and when infiltration reaches the field capacity, it yields the groundwater flow which is calculated with the amounts of infiltration and the distribution of the water storage capacity. For instance, the proposed model is applied to Dongwan Basin, a semi-humid region located at the middle reach of Yellow River. Results are also compared with those obtained by the Xinanjiang model. It turns out that the statistically-based runoff-yield model could achieve the promising results with acceptable accuracy for flood events' simulation and forecast.