Predicting gully erosion susceptibility in South Africa by integrating literature directives with regional spatial data

Abstract

Gully erosion has been identified as a severe land degradation process with environmental and socio-economic consequences. Identifying areas susceptible to gully erosion will aid in developing strategies to inhibit future degradation. Various approaches have been implemented to predict and map gully erosion susceptibility but are mostly restricted to small geographical extents because of process limitations. Here, we introduce a novel method that predicts gully erosion susceptibility on a regional/national scale (1.22 million km2) by synthesising literature directives with a statistical approach. Findings from a literature review were used to extract physiographic properties associated with gully erosion that was conditioned to characterise susceptibility by using the Frequency Ratio model. The conditioned physiographic properties were aggregated by a weighted overlay procedure using an aggregation of controlling factors derived from the literature review as a weighting system. The gully susceptibility index (GSI) model was validated against a published gully inventory map (n = 163 019) and randomly generated 1-km2 tessellation zones from which primary validation data were derived. Although uncertainties within the modelling procedure exist (e.g., gully site distribution, the spatial resolution of input data and determination of gully points), the validation shows that the GSI model is generally robust, identifying areas of contrasting susceptibilities. Furthermore, findings converge with other susceptibility metrics, which have been derived by different methodologies. Because empirical gully erosion research has been conducted worldwide, this model could be applied to regional-scale gully susceptibility modelling assessments (as a solitary method or combined with primary data) in other parts of the world. Additionally, the GSI model can be adopted to model environmental change scenarios.