Headcut retreat is one of the important processes of gully erosion, which leads to intensive sediment  yield and threaten the soil quality of a region (Oostwoud Wijdenes et al., 2000; Vandekerckhove et al., 2001; 2003). The waterfall at the headcuts dissipated flow kinetic energy to gully beds and formed plunge pools right under the headcuts (Stein et al., 1993). The development of plunge pools caused rill headcut migration to upstream drainage area (Bennett et al., 1999), and the shape factors were applied to predict the linear headcut retreat rates of rills (Alonso et al., 2002). Flores-Cervantes et al (2006) extended the model to hillslope gully heads in the model channel-hillslope integrated

landscape development (CHILD). Campo-Bescos et al (2013) calibrated the model in northeast Spain and found the differences between field observation and model simulation were only 0.05 m3 year-1. While the plunge pools were not typical in some gullies in fields, and how to predict the headcut erosion rates without shape factors of plunge pools needs further study. The aim of this study is to detect the relationships between headcut retreat rate and flow hydraulics and erosion processes of gully beds, and find out the key influencing factors on gully head eroison based on an in-situ scouring experiment on 5 simulated gully heads in Dry-hot valley in southwest China.