Rill erosion is known as the most important form of water erosion on agricultural land. Rill morphology plays a significant role in determining the runoff and soil loss from sloping farmland. However, few attempts have been done on the accurate description of rill throughout the study area for that eroding rill evolves morphologically in time and space. The objective of this study was to quantify the natural rill morphology under controlled laboratory conditions, aimed to provide more insight and detail information for later experiments. The soil used in this study was the loess soil with 28.3% sand, 58.1% silt, and 13.6% clay content from Ansai, Shaanxi province, China. It was collected from the top 20 cm in the A horizon of a well-drained site. The loess soil was packed in a soil box which is 10 m long, 3 m wide and 0.5 m deep, and performed to two sequent simulated rainfalls at an interval of 24 h. The slope was set as 15°. Each simulated rainfall lasted 65 minutes and the rainfall intensity was 60 mm/h. Rainfall intensity, rainfall uniformity and raindrop diameter were calibrated and reached the experimental requirements. In this study, the spatial distribution of rill erosion, rill development situation and relationships between the length away from rill head, rill width and depth were analyzed systematically. Results showed that discontinuous rills were widely distributed on the slope and headward erosion was the pre-dominant erosion manners in the first simulated rainfall. The accumulated length of all rills was 39.3 m and the total surface area reached up to 14.2% of the total soil box. However, the second simulated rainfall was mainly dominated by continuous rills, and the side-wall collapse of rills and fragmentation degrees of the slope strengthened. Meanwhile, the accumulated length increased by 32.1% and total surface area rose by 115.6%.Compared with the first simulated rainfall, average rill width, depth and rill erosion average depth all increased during the second run. And the increases occurred principally in the 4-7 m slope. The average rill width and depth of unit slope length increased firstly and decreased afterwards along with the increase of slope length. This indicated the growth and decline of runoff energy could affect the development processes and morphological characteristics of rill erosion. The length of each rill was less than 3 m and rill width increased with the length away from rill head for the first simulated rainfall. But the trend increased first and then decreased for the second simulated rainfall. Moreover, similar trends of rill depth developments were found for the two simulated rainfalls. The best regression relationship was found between rill width and depth. Rill depth increased with rill width at first, and then kept stable.