Abstract: Abstract: The soil denudation caused by concentrated water flow is the primary source of sediment in an eroded rill. Soil detachment rate of sediment loading flow is not only significant for understanding the dynamic process of rill erosion but also an important parameter in soil erosion prediction and soil erosion process modeling. Laboratory simulation is an important method to study rill erosion. Previous researches have suggested an experimental method for rill erosion process study. In their method sediment laden samples are taken from different rill lengths at different time. The rill erosion process along the simulated rill is integrated from the experimental data obtained from different rills lengths. Previous experimental approach is not only disrupted the continuity of rill erosion process but also time consuming. In order to get a better understand of rill erosion process, an improved approach was used for rill erosion process study. In this newly-suggested method, sediment laden samples were taken along the rill simultaneously and the length of the experimental flume was extended to 12 m to ensure sediment loading flow can reach its transport capacity under every hydraulic conditions. A flume platform that was 12 m long and 3 m wide was used as a base to construct a flume that was 12 m long, 0.6 m wide and 0.5 m deep. The flume was sub-divided into six rills that were 0.1 m wide and 12 m long using upright PVC boards to form well-defined rills. A typical silt-loam soil with 22.34% sand, 61.35% silt and 16.31% clay particles from the Loess Plateau of China was used in the experiment. The soil was air-dried and passed through an 8-mm sieve before packed into the simulated rills. The depth of soil in the flume was 25 cm. The bottom 5 cm of the soil was packed to a bulk density of 1 500 kg/m3 to imitate the plow pan layer. The top 20 cm was packed to bulk density of 1 200 kg/m3 as the experimental layer. The soil near the PVC boards was packed slightly higher than the middle to force the water to converge and to minimize the boundary effect. Prior to the experiment, the soil was saturated with a rainfall simulator and allowed to drain for 24 h to ensure homogeneous initial soil moisture content. The experiments were conducted with five slope gradients (5°, 10°, 15°, 20°, 25°) and three inflow rates (2, 4, 8 L/min) with three replicates. Sediment laden samples were simultaneously taken along the rill at the distances of 0.5, 1, 2, 3, 4, 5, 6, 7, 8, 10, and 12 m to determine the rill detachment rate along the rill. Three samples were taken at 1.0-minute intervals for each run. The results indicated that rill detachment rate decreased linearly with sediment concentration and exponentially with rill length (under slope gradients of 15°, 20° and 25°)for all flow rates and slope gradients. That meant the soil detachment rate and sediment concentration moved was opposite to each other. Detachment rate had its maximum value when the sediment concentration was at its minimum value. This variation trend was more obvious at steeper slope gradient and higher flow rate. Both slope gradient and flow rate had a positive influence on detachment rate. Additionally, analytic result indicated that rill detachment rate was a function of sediment concentration, rill length, flow rate and slope gradient. The detachment rate obtained using the traditional method and the ones from this study were compared. The correlation coefficient of the two data sets is 0.917 which indicated that detachment rate can be reliably estimated by the two methods. But the newly-suggested method was more precise, time efficient and easy to operate. The study provided a guideline for rill erosion prevention and establishment of rill erosion prediction model.