Abstract: Abstract: Soil surface roughness (SSR) is used to describe the surface spatial variation of elevation on rough slope and it plays a critical role in runoff generation and erosion process. Due to the presence of mounds and depressions distributed on rough slope, these micro-reliefs modify the surface depression storage capacity and extended rainwater retention time on slope. Consequently, it regulates the runoff generation and soil erosion process. Most of studies on SSR are focused on its measurement methods (Pin meter, Chain method, Laser scanner and Photogrammetry) and quantification index (Random Roughness, Mean Upslope Depression, Limiting Elevation Difference and Limiting Slope, etc.). And they also investigate SSR impacts on runoff and soil loss and how it evolves in soil erosion process. However, little information is available on the effect of soil surface roughness on runoff generation mechanism. Therefore, the aim of this study is to clarify how soil surface roughness impacts on runoff generation and investigate dominant impact on runoff generation process in sloping land under simulated rainfall. To achieve this purpose, the following specific objectives were addressed: 1) comparing the distribution of runoff generation points on rough and smooth slope; 2) analyzing the depression storage capacity, runoff generation time and process features during the simulated rainfall-runoff event for separating the direct and indirect impacts of roughness on runoff generation. Experiments were conducted on indoor soil box and outdoor runoff plots with four slope gradients (5?, 10?, 15?, 20?) under rainfall simulation condition. Five soil surface conditions were treated, included artificial digging (AD), artificial hoeing (AH), contour plough (CP), traditional plow (TP) and smooth surface (SS), by applied simulated at rainfall intensity of 60, 90, 120 mm/h. These results showed that the runoff generation point distance along the runoff downward direction on the rough and smooth surface was 12-181 cm, 40-180 cm, respectively. The coefficient variation of the distance at both slopes was 34.4%-52.1%, 15.15%-31.1%, respectively. It was indicated that the runoff generation points in rough surface were more dispersal compared with smooth surface. However, the depressions in rough slope may converge the overland flow, thus led to extend runoff retention time on slope and delay runoff generation time. The delayed impact would decrease as increasing rainfall intensity and slope gradient. The roughness affected the runoff generation time by two aspects. One was the surface depression storage, which called the direct impact. Another was increased infiltration water head through the depression storage, which called the indirect impact. The predicted time to runoff generation (Tp) was calculated by depression capacity (DSC, mm) and rainfall intensity (RI, mm/min) on a rough slope. Consequently, the Tp may reflect the direct impact on runoff generation in a rough surface. In order to separate the direct and indirect impact, the ratio of predicted time to runoff generation (Tp) and measured time to runoff initiation (Tr) was 2.2%-36.2%. It was showed that the indirect impact was the dominant process to delay runoff generation time, which means the depression storage improved slope infiltration capacity. The depressions in rough surface acted as a convergence role in runoff generation process. Therefore, these findings will improve the understanding of soil surface roughness impacts on runoff generation mechanism.