Modelling gully erosion for a small catchment on the Chinese Loess Plateau

The rolling hills region of the Chinese Loess Plateau is one of the areas with the highest erosion rates on earth. A striking feature of this area is the occurrence of many large, permanent gullies. A 3.5-km(2) catchment was selected to study the processes of erosion and to adapt the storm-based Limburg Soil Erosion Model (LISEM) to the conditions prevailing on the Loess Plateau. Part of this work consisted of mapping and measuring the largest gully headcuts. The amount of loose soil material beneath the headcuts was also estimated. Observations suggest that gully headcuts are relatively stable (i.e., do not migrate rapidly), but that gullies can nevertheless produce significant amounts of sediment during overland flow events. Erosion of headcuts occurs mainly by soil falls in between storms. The loose soil material produced by these soil falls accumulates on the gully bottom. As the LISEM simulates storm erosion, the development of gullies over time can be ignored, and only the amount of material produced by them during runoff events needs to be studied. A digital elevation model (DEM) was used to estimate the position of existing gully heads by applying an adapted form of the Montgomery and Dietrich [Science 255 (1992) 826] index. Using the assumption that headcuts are vertical, it is possible to calculate headcut height from, the slope angle map. A simple stability model, which assumes soil falls on gully headcuts to be a function of soil moisture content and headcut height, was applied. This daily-based model can then be used to simulate the accumulation of loose soil material below the headcut. The results show that while the DEM is not accurate enough to allow the detection of individual headcuts, this method can be used to produce a reasonable estimate of the amount of loose soil material available. A map showing the amount of loose soil material accumulated can then serve as input for a storm-based erosion model such as LISEM. (C) 2003 Elsevier Science B.V. All rights reserved.