Abstract: The current distributed unit hydrograph theory does not consider the time-varying routing paths due to variations in soil moisture content. To this end, a novel concept of dynamic routing path is proposed. First, a slope velocity formula considering both rainfall intensity as well as temporal and spatial distribution characteristics of soil moisture content is derived. Furthermore, the topographic index is introduced to describe the spatial distribution of watershed water storage capacity, and the flow velocity distribution field of the raster scale is evaluated. A collection of dynamic routing paths under various soil moisture contents are established. On this basis, the time-varying distributed unit hydrograph based on the dynamic flow routing paths is deduced. The Longhuwei and Dongshi River basins are selected as two case studies. The SCS model is employed for the runoff generation calculation, and the proposed method is then adopted for the flow routing calculation. The multi-objective method covering low flow error, high flow error, and flood error are introduced to optimize the calibrated parameters. The simulation results of 28 floods reveal that the relative error of the flood peak is within ±15%, the error of peak appearance time is within ±6 h, and the average value of the Nash efficiency coefficient exceeds 0.8. Compared with the current approach, the proposed method can provide a more accurate flow routing time distribution field, and the accuracy of flood prediction is chiefly enhanced.