As an important parameter in the soil erosion model, soil surface roughness(SSR) is used to quantitatively describe the micro-relief on agricultural land. SSR has been extensively studied both experimentally and theoretically; however, no studies have focused on understanding SSR on the Loess Plateau of China. This study investigated changes in SSR for three different tillage practices on the Loess Plateau of China and the effects of SSR on runoff and erosion yield during simulated rainfall. The tillage practices used were zero tillage(ZT), shallow hoeing(SH) and contour ploughing(CP). Two rainfall intensities were applied, and three stages of water erosion processes(splash erosion(I), sheet erosion(II) and rill erosion(III)) were analyzed for each rainfall intensity. The chain method was used to measure changes in SSR both initially and after each stage of rainfall. A splash board was used to measure the splash erosion at stage I. Runoff and sediment data were collected continuously at 2-min intervals during rainfall erosion stages II and III. We found that SSR of the tilled surfaces ranged from 1.0% to 21.9% under the three tillage practices, and the order of the initial SSR for the three treatments was ZT0.49) for the ZT, SH and CP treatments. The runoff initiation time was longer in the SH and CP treatments than in the ZT treatment. There were no significant differences in the total runoff yields among the ZT, SH and CP treatments. Sediment loss was significantly smaller(P<0.05) in the SH and CP treatments than in the ZT treatment. As an important parameter in the soil erosion model, soil surface roughness (SSR) is used to quantitatively describe the micro-relief on agricultural land. However, no, studies have focused on understanding SSR on the Loess Plateau of China. This study analyzed changes in SSR for three different tillage practices on the Loess Plateau of China and the effects of SSR on runoff and erosion yield on simulated rainfall. The tillage practices used were zero tillage (ZT), shallow hoeing ( SH) and contour plowing (CP). Two rainfall intensities were applied, and three stages of water erosion processes (splash erosion (I), sheet erosion (II) and rill erosion method was used to measure changes in SSR both initially and after each stage of rainfall. A splash board was used to measure the splash erosion at stage I. Runoff and sediment data were collected continuously a t 2-min intervals during rainfall erosion stages II and III. We found that SSR of the tilled surfaces ranged from 1.0% to 21.9% under the three tillage practices, and the order of the initial SSR for the three treatments was ZT 0.49) for the ZT, SH and CP treatments. The runoff initiation time was longer in the SH and CP treatments than in the ZT treatment. There were no significant differences in the total runoff yields among the ZT, SH and CP treatments. Sediment loss was significantly smaller (P <0.05) in the SH and CP treatments than in the ZT treatment.