Rill erosion dynamics in smallholder farming systems of wet tropical Africa

Tropical Africa is globally one of the most sensitive regions to accelerated soil erosion, and much of its cropland is characterised by a substantial yield gap. In particular, the White Nile-Congo ridge (NiCo) region of the eastern Democratic Republic of the Congo (DR Congo) and Uganda is a hotspot for issues relating to food security driven by soil degradation due to steep terrain, highly erosive rainfall and low soil cover. Most soil erosion studies in the region are based on plot or large-scale modelling. Both approaches lack information on inter-field connectivity processes, which are especially important in smallholder farming systems with average field sizes below 0.1 ha. To address this knowledge gap, an unmanned aerial vehicle (UAV) based high-spatial and temporal-resolution monitoring campaign was carried out over four smallholder farming areas within the eastern DR Congo and western Uganda, with substantial differences in cropland management and productivity. The campaign covered 833 individual fields, which were monitored up to twice per month (for two years) using UAV-based aerial photography to provide insight into event-based rill erosion processes and landscape connectivity. The aerial photography data were classified according to field conditions: (i) dense vegetation cover, (ii) low vegetation cover or bare soil without signs of rill erosion, (iii) low vegetation cover or bare soil with signs of rill erosion. Land-use patchiness associated with smallholder farming systems was found to reduce inter-field connectivity and to promote highly localised rill development. Furthermore, rill erosion in the NiCo region is not an episodic process but takes place regularly during the rainy season due to frequent storm events falling on bare soil in fields left fallow for individual cultivation periods. Soil erosion dynamics of smallholder farming regions in the study area, therefore, pose unresolved challenges regarding their implementation in large-scale predictions.