Abstract
Gully erosion is a worldwide matter of concern because of the irreversible losses of fertile land, which often have severe environmental, economic and social consequences. While most of the studies on the gullying process have investigated the involved mechanisms (either overland flow incision, seepage or piping erosion), only few have been conducted on the controlling factors of gully wall retreat, an important, if not the dominant, land degradation process and sediment source in river systems. In a representative 4.4 km2 degraded area of the Drakensberg foothills (South Africa) the main objective of this study was to evaluate the relationship between the rate of gully bank retreat (GBR) and parent material, soil types and selected terrain attributes (elevation, specific drainage area, mean slope gradient, slope length factor, stream power index, compound topographic index and slope curvatures). The survey of gully bank retreat was performed during an entire hydrological year, from September 2007 to September 2008, using a network of pins (n = 440 from 110 pits). Both the gully contours and pin coordinates were determined, using a GPS with a 0.5 m horizontal accuracy (n = 20,120). The information on the parent material and the soil types was obtained from field observations complemented by laboratory analysis, while terrain attributes were extracted from a 20 m DEM generated from 5 m interval contour lines. The average GBR value for the 6512 m of gully banks found in the area was 0.049 ± 0.0013 m y− 1, which, considering bank height and soil bulk density, corresponded to an erosion rate of 2.30 ton ha− 1 y− 1. There was no significant difference in GBR between sandstone and dolerite and between Acrisols and Luvisols. Despite a weak one-to-one correlation with the selected terrain attributes (r < 0.2), a principal component analysis (PCA), the first two axes of which explained 68% of the data variability, pointed out that GBR was the highest at hillslope inflexion points (profile and plan slope curvatures close to zero), in the vicinity of the head cuts and for drainage areas up to 500 m2, as both situations experience a high removal rate of the soil material produced from the gully bank collapse and protecting gullies from laterally retreating. These results could be used to digitally map the more active gully banks for the improved implementation of preventive measures of gully growth, if high resolution DEMs are available. There remained, however, a certain amount of unexplained variability in the data, that further research studies on the mechanisms and associated factors of control of GBR could help to address