Towards intercrop ideotypes: non-random trait assembly can promote overyielding and stability of species proportion in simulated legume-based mixtures

Backgrounds and Aims: A major challenge when supporting the development of intercropping systems remains the design of efficient species mixtures. The ecological processes that sustain overyielding of legume-based mixtures compared to pure crops are well known. but their links to plant traits remain to be unravelled. A common assumption is that enhancing trait divergence among species for resource acquisition when assembling plant mixtures should increase species complementarity and improve community performance. Methods: The Virtual Grassland model was used to assess how divergence in trait values between species on four physiological functions (namely light and mineral N acquisition, temporal development, and C-N use efficiency) affected overyielding and mixture stability in legume-based binary mixtures. A first step allowed us to identify the model parameters that were most important to interspecies competition. A second step involved testing the impact of convergent and divergent parameter (or trait) values between species on virtual mixture performance. Results: Maximal overyielding was achieved in cases where trait values were divergent for the physiological functions controlling N acquisition and temporal development but convergent for light interception. It was also found that trait divergence should not affect competitive abilities of legume and non-legumes at random. Indeed, random trait combinations frequently led to reduced mixture yields when compared to a perfectly convergent neutral model. Combinations with the highest overyielding also tended to be associated with mixture instability and decreasing legume biomass proportion. Achieving both high overyielding and mixture stability was only found to be possible under low or moderate N levels, using combinations of traits adapted to each environment. Conclusions: No simple assembly rule based on trait divergence could be confirmed. Plant models able to infer plant-plant interactions can be helpful for the identification of major interaction traits and the definition of ideotypes adapted to a targeted intercropping system.