Low but highly variable soil organic carbon stocks across deeply weathered eroding and depositional tropical landforms

Abstract

Soil organic carbon (SOC) dynamics in temperate regions are highly affected by lateral soil fluxes induced by soil erosion. SOC dynamics in eroding tropical cropland systems characterized by deeply weathered soils and heterogeneous small-scale subsistence farming structures, however, are not well understood yet. Along topographic gradients in the East African Albertine rift region, we investigated the differences in SOC stocks and persistence for the upper meter of tropical soils developed from geochemically distinct parent materials and cultivated by subsistence farmers. We show that SOC stocks and persistence do not follow topography-driven patterns expected from research on less weathered, more fertile soils of temperate climate zones and more large-scale farming systems. At all investigated topographic positions, the SOC stocks were low compared to temperate regions while variability of stocks was high in both top- and subsoil. Full profile (0 – 100 cm) SOC stocks ranged from 256.1 t C ha^-1 to 297.3 t C ha^-1 at plateaus, from 224.4 t C ha^-1 to 276.1 t C ha^-1 at slopes and from 305.1 t C ha^-1 to 366.0 t C ha^-1 at footslopes. Independent of soil parent material and unless situated on very steep slopes (>15 % slope steepness), SOC stocks in eroding positions are, therefore, similar to those in non-eroding landscape positions and stable at a low level despite heavy erosion. Our results further suggest that deposition of eroded topsoil material at footslopes only slightly increases SOC stocks. Therefore, SOC stocks in this rapidly eroding tropical systems seem not to be heavily altered by soil redistribution while other soil parameters indicate heavy signs of soil disturbance. Tropical soil features and the distinctiveness of small scale subsistence farming practices create an extremely patchy and variable distribution of SOC, which requires us to rethink the way these landscapes can be modelled to represent C dynamics.