Abstract: Abstract: Recently, infrastructure construction developed rapidly and produced a large amount of engineering accumulation, which becomes a new critical point of soil and water loss. Particularly, rills formed quickly on the slope of engineering accumulation under high rainfall intensity and triggered intense soil erosion. Therefore, it is essential to study on the dynamic process of rill development and its impact on runoff and sediment yield. In this study, scouring-erosion experiments were conducted on steep slopes of engineering accumulation under different slope gradients (24°, 28° and 32°) with different flow discharge (5, 9, 13 and 17 L/min), simulating the surface runoff process under the rainfall intensity of 0.5, 1.0, 1.5 and 2.0 mm/min. The zonal soil type in test area is Heilu soil. The experimental plots were constructed in engineering excavated slope, meanwhile, soil parent material was the Loess of Malan at first and then turned into spoil derived from construction sites. The spoil here was sandy loam, with a soil-stone ratio more than 9:1 and the size of particles <1 mm. Each plot was 20 m long and 5 m wide, with 0.5 m deep soil layer and without vegetation. It was used to simulate the dumping engineering accumulation of gully region in loess plateau. According to the experiment design, runoff plot was divided into 5 one-meter-wide smaller plots by iron sheets, one of which was rebuilt as steps. To ensure the same original background, four rill experiment plots were backfilled and compacted after last experimental events. Because the randomness of the slope rill development may cause much difficulty on continuous observation, the slope surface was slightly concavely shaped to form a single rill. The morphological parameters such as rill width, rill depth, width-depth ratio and sectional area were selected to reveal the dynamic change of rill shape during erosion processes. The results showed: 1) The rill width and the rill depth both rapidly increased within the first 9 mins. The rill width accounted to 57%-90% of the ultimate width and the rill depth accounted to 38%-73% of the final depth.；2) With the scouring time continuing, the width-depth ratio first decreased and then showed a stable trend. Within the starting 27 mins, the width-depth ratio rapidly decreased and indicated the attenuated ability of rill development along the lengthways slope, finally, the ratio was stable between 0.81 and 1.48, with the ultimate rill section presented as a rectangular shape. 3) The rill width and rill depth had a positive correlation with flow intensity and flow intensity had a more significant effect on rill development compared to slope gradients. 4) The rill width and rill depth showed logarithmic relationships with discharge time; the sectional area had a linear relationship with discharge time; the rill width had an exponential relationship with runoff; the rill depth had a power function relationship with runoff; the sectional area had power function relationships with both cumulative sediment yield and cumulative runoff. Temporal variability existed in rill morphological development on slope. As a whole, sectional area described the dynamic process of erosion and the width-depth ratio was an important indicator to reveal the characteristics of rill morphological development and erosive ability. This research may provide some theoretical reference for quantifying dynamic changes of rill morphology index on slopes of engineering accumulation.