Simulating medium-term effects of cropping system diversification on soil fertility and crop productivity in southern Africa.

• APSIM tested against six and eight-year crop rotation trials in southern Africa. • Effect of diversification on average productivity, yield trends, soil carbon and nitrogen properties investigated. • Low input systems benefited from crop diversification that included legumes through increased Nmin content. • Commercial systems faced a productivity-environment trade-off in the short term, but intercropping seemed promising. Diversification of cropping is perceived as a strategy to simultaneously achieve high productivity and maintain environmental sustainability. In southern Africa, however, due to a lack of medium to long-term field trials, there is missing quantitative information. Utilising the capability of agro-ecosystem models to quantify the interactions of crop productivity with management and environmental variables, the APSIM model was evaluated against six and an eight-year field trial datasets comprised of different crop rotations and fertiliser rates under two contrasting agro-ecological conditions in South Africa (clay soil with a mean rainfall of 871 mm versus a sandy soil with a 570 mm rainfall). Model output was compared to observed grain yield, aboveground dry matter, soil organic carbon (SOC), and mineral nitrogen (N min) dynamics. APSIM was able to reproduce the observed grain yield dynamics fairly well as indicated by a Wilmott index of agreement of 0.90. Prediction accuracy, as indicated by the absolute model error across all crops, however, only reached 39%. Simulated N min and SOC dynamics showed similar patterns to the observations. Subsequently, the model was applied in a ten-year simulation experiment with rotation treatments (14 rotations, respectively intercropping systems, including a maize monoculture control), fertiliser levels (zero and 70 kg N ha−1), and residue management (retained and removed) for the two sites. For low input systems, such as smallholder farms, residue management and legume integration are of the utmost importance to maintain SOC and more pronounced Nmin levels, which, for the sandy soil, resulted in an average maize yield increase of up to 1000 kg ha−1. Maize monoculture treatments with residues removed reduced SOC moderately by 0.04–0.08 %, while yields declined strongly (>1000 kg ha−1) over the simulated period of ten years. In commercial, fertilised cropping systems, allocating land to cultivate crops other than maize reduced the simulated total yield performance. This diversification disadvantage has to be considered against the benefits of increased SOC and yields in the medium-term, i.e. a period of ten years. For the commercial systems, maize intercropped with delayed sown oats or cowpea appeared promising. [ABSTRACT FROM AUTHOR]