Abstract：Gully headcut erosion become the main cause for the longitudinal and horizontal surface, surface fragmentation and area shrinkage in the gully region of the Loess Plateau in China,threatening the development of local agriculture, economy and ecological environment. At present, the vegetation hasplayed an important role in controlling the gully headcut erosion in this region. Most of the previous studies mainly focused on the influence of the above-ground parts of the vegetation on the basin and even the regional scale.However, little is known about the effects of the root density of vegetation on the gully headcut erosion. The effects of sediment production and its morphological evolution are also lacking, particularly on the loess. Therefore, a combined experiment of simulated rainfall and runoff scouring was conducted to explore the effects of the grass (Agropyron cristatum) root densities on the erosion process and morphological evolution during the gully headcut erosion. The experimental plot consist of the upstream area, gully head and gully bed. The length and width of the upstream area were 8.0 m and 1.5 m. The gully headwall height was 1.2 m. The length and width of the gully bed were 1.0 m and 1.5 m. The slope of the upstream area was consistent with that of the gully bed, being set as 3°.Compared to the bare land experiment area, this paper presents the process of soil loss, gully head retreat, gully cutting and development, in the different root density around the experiment area of Agropyron cristatum (planting space: 20 cm×20 cm, C1; 15 cm×15 cm, C2; 10 cm×10 cm, C3). The results showed that: 1) Compared with the control plot, the sediment yield of grass-cover plot (C1-C3) decreased by 64.32%, 70.31%and 69.92%, respectively. When the plant spacing of the Agropyron cristatum plants was 15 cm × 15 cm, the sand reduction benefit was the largest. 2) In the control plot, the gully headcut erosion mainly includedthe following steps: the upstream flow incision, headwall erosion by the wall flow, plunge pool erosion by the jet flow, and finally the gully bank collapse,whereas,in the grass-covered plot,the gully headcut erosion was determined by the wall flow, plunge pool erosion by the jet flow, and the collapse of the hanging soil-root matrix at the gully head. The collapse was also the main reason for the gully headcut erosion of grassland. Alternatively, the accumulation distance of gully head in the grass-covered plot was shortened by 75.61% and 78.87%, and the accumulation distance of gully head inboth plots showed a significant power function with time. 3) The longitudinal section of erosion gully in the control plot was in ladder-shaped, and the gully head was rectangular; while in the grass-covered plot was trapezoid and arc-shaped. The average incision depth of the gully channel in the grass-covered was 1.64-1.92 times as deep as that in the control plot. However, the developed area of the gully channel decreased with the increase of root density, and the area of the gully in the grass-covered plot decreased by 68.0%- 74.0% compared with that in the control plot. The findings can provide a promisingapproach for the implementation of the gully stabilization and loess-tableland protection in the gully region of Loess Plateau in China